Server IP : 66.29.132.122 / Your IP : 13.59.83.52 Web Server : LiteSpeed System : Linux business142.web-hosting.com 4.18.0-553.lve.el8.x86_64 #1 SMP Mon May 27 15:27:34 UTC 2024 x86_64 User : admazpex ( 531) PHP Version : 7.2.34 Disable Function : NONE MySQL : OFF | cURL : ON | WGET : ON | Perl : ON | Python : ON | Sudo : OFF | Pkexec : OFF Directory : /opt/alt/ruby30/share/gems/gems/rdoc-6.3.4.1/lib/rdoc/ |
Upload File : |
# coding: US-ASCII # frozen_string_literal: true ## # This class is a wrapper around File IO and Encoding that helps RDoc load # files and convert them to the correct encoding. module RDoc::Encoding HEADER_REGEXP = /^ (?: \A\#!.*\n | ^\#\s+frozen[-_]string[-_]literal[=:].+\n | ^\#[^\n]+\b(?:en)?coding[=:]\s*(?<name>[^\s;]+).*\n | <\?xml[^?]*encoding=(?<quote>["'])(?<name>.*?)\k<quote>.*\n )+ /xi # :nodoc: ## # Reads the contents of +filename+ and handles any encoding directives in # the file. # # The content will be converted to the +encoding+. If the file cannot be # converted a warning will be printed and nil will be returned. # # If +force_transcode+ is true the document will be transcoded and any # unknown character in the target encoding will be replaced with '?' def self.read_file filename, encoding, force_transcode = false content = File.open filename, "rb" do |f| f.read end content.gsub!("\r\n", "\n") if RUBY_PLATFORM =~ /mswin|mingw/ utf8 = content.sub!(/\A\xef\xbb\xbf/, '') enc = RDoc::Encoding.detect_encoding content content = RDoc::Encoding.change_encoding content, enc if enc begin encoding ||= Encoding.default_external orig_encoding = content.encoding if not orig_encoding.ascii_compatible? then content = content.encode encoding elsif utf8 then content = RDoc::Encoding.change_encoding content, Encoding::UTF_8 content = content.encode encoding else # assume the content is in our output encoding content = RDoc::Encoding.change_encoding content, encoding end unless content.valid_encoding? then # revert and try to transcode content = RDoc::Encoding.change_encoding content, orig_encoding content = content.encode encoding end unless content.valid_encoding? then warn "unable to convert #{filename} to #{encoding}, skipping" content = nil end rescue Encoding::InvalidByteSequenceError, Encoding::UndefinedConversionError => e if force_transcode then content = RDoc::Encoding.change_encoding content, orig_encoding content = content.encode(encoding, :invalid => :replace, :undef => :replace, :replace => '?') return content else warn "unable to convert #{e.message} for #{filename}, skipping" return nil end end content rescue ArgumentError => e raise unless e.message =~ /unknown encoding name - (.*)/ warn "unknown encoding name \"#{$1}\" for #{filename}, skipping" nil rescue Errno::EISDIR, Errno::ENOENT nil end def self.remove_frozen_string_literal string string =~ /\A(?:#!.*\n)?(.*\n)/ first_line = $1 if first_line =~ /\A# +frozen[-_]string[-_]literal[=:].+$/i string = string.sub first_line, '' end string end ## # Detects the encoding of +string+ based on the magic comment def self.detect_encoding string result = HEADER_REGEXP.match string name = result && result[:name] name ? Encoding.find(name) : nil end ## # Removes magic comments and shebang def self.remove_magic_comment string string.sub HEADER_REGEXP do |s| s.gsub(/[^\n]/, '') end end ## # Changes encoding based on +encoding+ without converting and returns new # string def self.change_encoding text, encoding if text.kind_of? RDoc::Comment text.encode! encoding else # TODO: Remove this condition after Ruby 2.2 EOL if RUBY_VERSION < '2.3.0' text.force_encoding encoding else String.new text, encoding: encoding end end end end