module Bringit::EncodingHelper

Constants

ENCODING_CONFIDENCE_THRESHOLD

This threshold is carefully tweaked to prevent usage of encodings detected by CharlockHolmes with low confidence. If CharlockHolmes confidence is low, we're better off sticking with utf8 encoding. Reason: git diff can return strings with invalid utf8 byte sequences if it truncates a diff in the middle of a multibyte character. In this case CharlockHolmes will try to guess the encoding and will likely suggest an obscure encoding with low confidence. There is a lot more info with this merge request: gitlab.com/gitlab-org/gitlab_git/merge_requests/77#note_4754193

Public Instance Methods

encode!(message) click to toggle source
# File lib/bringit/encoding_helper.rb, line 16
def encode!(message)
  return nil unless message.respond_to? :force_encoding

  # if message is utf-8 encoding, just return it
  message.force_encoding("UTF-8")
  return message if message.valid_encoding?

  # return message if message type is binary
  detect = CharlockHolmes::EncodingDetector.detect(message)
  return message.force_encoding("BINARY") if detect && detect[:type] == :binary

  # force detected encoding if we have sufficient confidence.
  if detect && detect[:encoding] && detect[:confidence] > ENCODING_CONFIDENCE_THRESHOLD
    message.force_encoding(detect[:encoding])
  end

  # encode and clean the bad chars
  message.replace clean(message)
rescue
  encoding = detect ? detect[:encoding] : "unknown"
  "--broken encoding: #{encoding}"
end
encode_utf8(message) click to toggle source
# File lib/bringit/encoding_helper.rb, line 39
def encode_utf8(message)
  detect = CharlockHolmes::EncodingDetector.detect(message)
  if detect
    CharlockHolmes::Converter.convert(message, detect[:encoding], 'UTF-8')
  else
    clean(message)
  end
end

Private Instance Methods

clean(message) click to toggle source
# File lib/bringit/encoding_helper.rb, line 50
def clean(message)
  message.encode("UTF-16BE", undef: :replace, invalid: :replace, replace: "")
    .encode("UTF-8")
    .gsub("\0".encode("UTF-8"), "")
end