gitlab-org--gitlab-foss/lib/gitlab/blob_helper.rb

Ignoring revisions in .git-blame-ignore-revs. Click here to bypass and see the normal blame view.

162 lines
4.3 KiB
Ruby
Raw Normal View History

# frozen_string_literal: true
# This has been extracted from https://github.com/github/linguist/blob/master/lib/linguist/blob_helper.rb
module Gitlab
module BlobHelper
include Gitlab::Utils::StrongMemoize
def extname
File.extname(name.to_s)
end
def known_extension?
LanguageData.extensions.include?(extname)
end
def viewable?
!large? && text_in_repo?
end
MEGABYTE = 1024 * 1024
def large?
size.to_i > MEGABYTE
end
def binary_in_repo?
# Large blobs aren't even loaded into memory
if data.nil?
true
# Treat blank files as text
elsif data == ""
false
# Charlock doesn't know what to think
elsif encoding.nil?
true
# If Charlock says its binary
else
find_encoding[:type] == :binary
end
end
def text_in_repo?
!binary_in_repo?
end
def image?
['.png', '.jpg', '.jpeg', '.gif', '.svg', '.webp'].include?(extname.downcase)
end
# Internal: Lookup mime type for extension.
#
# Returns a MIME::Type
# rubocop:disable Gitlab/ModuleWithInstanceVariables
def _mime_type
if defined? @_mime_type
@_mime_type
else
guesses = ::MIME::Types.type_for(extname.to_s)
# Prefer text mime types over binary
@_mime_type = guesses.detect { |type| type.ascii? } || guesses.first
end
end
# rubocop:enable Gitlab/ModuleWithInstanceVariables
# Public: Get the actual blob mime type
#
# Examples
#
# # => 'text/plain'
# # => 'text/html'
#
# Returns a mime type String.
def mime_type
_mime_type ? _mime_type.to_s : 'text/plain'
end
def binary_mime_type?
_mime_type ? _mime_type.binary? : false
end
def lines
@lines ||=
if viewable? && data
# `data` is usually encoded as ASCII-8BIT even when the content has
# been detected as a different encoding. However, we are not allowed
# to change the encoding of `data` because we've made the implicit
# guarantee that each entry in `lines` is encoded the same way as
# `data`.
#
# Instead, we re-encode each possible newline sequence as the
# detected encoding, then force them back to the encoding of `data`
# (usually a binary encoding like ASCII-8BIT). This means that the
# byte sequence will match how newlines are likely encoded in the
# file, but we don't have to change the encoding of `data` as far as
# Ruby is concerned. This allows us to correctly parse out each line
# without changing the encoding of `data`, and
# also--importantly--without having to duplicate many (potentially
# large) strings.
begin
data.split(encoded_newlines_re, -1)
rescue Encoding::ConverterNotFoundError
# The data is not splittable in the detected encoding. Assume it's
# one big line.
[data]
end
else
[]
end
end
def content_type
# rubocop:disable Style/MultilineTernaryOperator
# rubocop:disable Style/NestedTernaryOperator
@content_type ||= binary_mime_type? || binary_in_repo? ? mime_type :
(encoding ? "text/plain; charset=#{encoding.downcase}" : "text/plain")
# rubocop:enable Style/NestedTernaryOperator
# rubocop:enable Style/MultilineTernaryOperator
end
def encoded_newlines_re
strong_memoize(:encoded_newlines_re) do
newlines = ["\r\n", "\r", "\n"]
data_encoding = data&.encoding
if ruby_encoding && data_encoding
newlines.map! do |nl|
nl.encode(ruby_encoding, "ASCII-8BIT").force_encoding(data_encoding)
end
end
Regexp.union(newlines)
end
end
def ruby_encoding
if hash = find_encoding
hash[:ruby_encoding]
end
end
def encoding
if hash = find_encoding
hash[:encoding]
end
end
def empty?
data.nil? || data == ""
end
private
def find_encoding
@find_encoding ||= Gitlab::EncodingHelper.detect_encoding(data) if data # rubocop:disable Gitlab/ModuleWithInstanceVariables
end
end
end