debian-mirror-gitlab/lib/gitlab/encoding_helper.rb

104 lines
3.5 KiB
Ruby
Raw Normal View History

2018-12-13 13:39:08 +05:30
# frozen_string_literal: true
2017-09-10 17:25:29 +05:30
module Gitlab
module EncodingHelper
extend self
# 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:
# https://gitlab.com/gitlab-org/gitlab_git/merge_requests/77#note_4754193
ENCODING_CONFIDENCE_THRESHOLD = 50
def encode!(message)
2018-03-17 18:26:18 +05:30
message = force_encode_utf8(message)
2017-09-10 17:25:29 +05:30
return message if message.valid_encoding?
# return message if message type is binary
detect = CharlockHolmes::EncodingDetector.detect(message)
2018-03-17 18:26:18 +05:30
return message.force_encoding("BINARY") if detect_binary?(message, detect)
2017-09-10 17:25:29 +05:30
if detect && detect[:encoding] && detect[:confidence] > ENCODING_CONFIDENCE_THRESHOLD
2018-03-17 18:26:18 +05:30
# force detected encoding if we have sufficient confidence.
2017-09-10 17:25:29 +05:30
message.force_encoding(detect[:encoding])
end
# encode and clean the bad chars
message.replace clean(message)
2018-03-27 19:54:05 +05:30
rescue ArgumentError => e
return unless e.message.include?('unknown encoding name')
2017-09-10 17:25:29 +05:30
encoding = detect ? detect[:encoding] : "unknown"
"--broken encoding: #{encoding}"
end
2018-03-17 18:26:18 +05:30
def detect_binary?(data, detect = nil)
detect ||= CharlockHolmes::EncodingDetector.detect(data)
detect && detect[:type] == :binary && detect[:confidence] == 100
end
def detect_libgit2_binary?(data)
# EncodingDetector checks the first 1024 * 1024 bytes for NUL byte, libgit2 checks
# only the first 8000 (https://github.com/libgit2/libgit2/blob/2ed855a9e8f9af211e7274021c2264e600c0f86b/src/filter.h#L15),
# which is what we use below to keep a consistent behavior.
detect = CharlockHolmes::EncodingDetector.new(8000).detect(data)
detect && detect[:type] == :binary
end
2017-09-10 17:25:29 +05:30
def encode_utf8(message)
2018-03-17 18:26:18 +05:30
message = force_encode_utf8(message)
return message if message.valid_encoding?
2017-09-10 17:25:29 +05:30
detect = CharlockHolmes::EncodingDetector.detect(message)
if detect && detect[:encoding]
begin
CharlockHolmes::Converter.convert(message, detect[:encoding], 'UTF-8')
rescue ArgumentError => e
Rails.logger.warn("Ignoring error converting #{detect[:encoding]} into UTF8: #{e.message}")
''
end
else
clean(message)
end
2018-03-17 18:26:18 +05:30
rescue ArgumentError
2018-11-18 11:00:15 +05:30
nil
2018-03-17 18:26:18 +05:30
end
2018-11-18 11:00:15 +05:30
def encode_binary(str)
return "" if str.nil?
2018-03-17 18:26:18 +05:30
2018-11-18 11:00:15 +05:30
str.dup.force_encoding(Encoding::ASCII_8BIT)
2018-03-17 18:26:18 +05:30
end
2019-05-18 00:54:41 +05:30
def binary_io(str_or_io)
io = str_or_io.to_io.dup if str_or_io.respond_to?(:to_io)
io ||= StringIO.new(str_or_io.to_s.freeze)
io.tap { |io| io.set_encoding(Encoding::ASCII_8BIT) }
2017-09-10 17:25:29 +05:30
end
private
2018-03-17 18:26:18 +05:30
def force_encode_utf8(message)
raise ArgumentError unless message.respond_to?(:force_encoding)
return message if message.encoding == Encoding::UTF_8 && message.valid_encoding?
message = message.dup if message.respond_to?(:frozen?) && message.frozen?
message.force_encoding("UTF-8")
end
2017-09-10 17:25:29 +05:30
def clean(message)
2018-04-04 21:44:52 +05:30
message.encode("UTF-16BE", undef: :replace, invalid: :replace, replace: "".encode("UTF-16BE"))
2017-09-10 17:25:29 +05:30
.encode("UTF-8")
.gsub("\0".encode("UTF-8"), "")
end
end
end