From f238d495b70a264abdb864fe8107e02766b285b4 Mon Sep 17 00:00:00 2001 From: Michael Koziarski Date: Tue, 3 Oct 2006 23:45:32 +0000 Subject: Add ActiveSupport::Multibyte. Provides String#chars which lets you deal with strings as a sequence of chars, not of bytes. Closes #6242 [Julian Tarkhanov, Manfred Stienstra & Jan Behrens] git-svn-id: http://svn-commit.rubyonrails.org/rails/trunk@5223 5ecf4fe2-1ee6-0310-87b1-e25e094e27de --- .../lib/active_support/core_ext/string.rb | 2 + .../lib/active_support/core_ext/string/unicode.rb | 42 ++ activesupport/lib/active_support/multibyte.rb | 7 + .../lib/active_support/multibyte/chars.rb | 129 ++++++ .../multibyte/generators/generate_tables.rb | 141 +++++++ .../multibyte/handlers/passthru_handler.rb | 9 + .../multibyte/handlers/utf8_handler.rb | 447 +++++++++++++++++++++ .../multibyte/handlers/utf8_handler_proc.rb | 44 ++ .../active_support/values/unicode_tables-1.8.4.dat | Bin 0 -> 654866 bytes .../active_support/values/unicode_tables-1.8.5.dat | Bin 0 -> 654866 bytes 10 files changed, 821 insertions(+) create mode 100644 activesupport/lib/active_support/core_ext/string/unicode.rb create mode 100644 activesupport/lib/active_support/multibyte.rb create mode 100644 activesupport/lib/active_support/multibyte/chars.rb create mode 100644 activesupport/lib/active_support/multibyte/generators/generate_tables.rb create mode 100644 activesupport/lib/active_support/multibyte/handlers/passthru_handler.rb create mode 100644 activesupport/lib/active_support/multibyte/handlers/utf8_handler.rb create mode 100644 activesupport/lib/active_support/multibyte/handlers/utf8_handler_proc.rb create mode 100644 activesupport/lib/active_support/values/unicode_tables-1.8.4.dat create mode 100644 activesupport/lib/active_support/values/unicode_tables-1.8.5.dat (limited to 'activesupport/lib/active_support') diff --git a/activesupport/lib/active_support/core_ext/string.rb b/activesupport/lib/active_support/core_ext/string.rb index 240e1ff1da..b9f075522f 100644 --- a/activesupport/lib/active_support/core_ext/string.rb +++ b/activesupport/lib/active_support/core_ext/string.rb @@ -3,6 +3,7 @@ require File.dirname(__FILE__) + '/string/conversions' require File.dirname(__FILE__) + '/string/access' require File.dirname(__FILE__) + '/string/starts_ends_with' require File.dirname(__FILE__) + '/string/iterators' +require File.dirname(__FILE__) + '/string/unicode' class String #:nodoc: include ActiveSupport::CoreExtensions::String::Access @@ -10,4 +11,5 @@ class String #:nodoc: include ActiveSupport::CoreExtensions::String::Inflections include ActiveSupport::CoreExtensions::String::StartsEndsWith include ActiveSupport::CoreExtensions::String::Iterators + include ActiveSupport::CoreExtensions::String::Unicode end diff --git a/activesupport/lib/active_support/core_ext/string/unicode.rb b/activesupport/lib/active_support/core_ext/string/unicode.rb new file mode 100644 index 0000000000..0e4c76fb86 --- /dev/null +++ b/activesupport/lib/active_support/core_ext/string/unicode.rb @@ -0,0 +1,42 @@ +module ActiveSupport #:nodoc: + module CoreExtensions #:nodoc: + module String #:nodoc: + # Define methods for handeling unicode data. + module Unicode + # +chars+ is a Unicode safe proxy for string methods. It creates and returns an instance of the + # ActiveSupport::Multibyte::Chars class which encapsulates the original string. A Unicode safe version of all + # the String methods are defined on this proxy class. Undefined methods are forwarded to String, so all of the + # string overrides can also be called through the +chars+ proxy. + # + # name = 'Claus Müller' + # name.reverse #=> "rell??M sualC" + # name.length #=> 13 + # + # name.chars.reverse.to_s #=> "rellüM sualC" + # name.chars.length #=> 12 + # + # + # All the methods on the chars proxy which normally return a string will return a Chars object. This allows + # method chaining on the result of any of these methods. + # + # name.chars.reverse.length #=> 12 + # + # The Char object tries to be as interchangeable with String objects as possible: sorting and comparing between + # String and Char work like expected. The bang! methods change the internal string representation in the Chars + # object. Interoperability problems can be resolved easily with a +to_s+ call. + # + # For more information about the methods defined on the Chars proxy see ActiveSupport::Multibyte::Chars and + # ActiveSupport::Multibyte::Handlers::UTF8Handler + def chars + ActiveSupport::Multibyte::Chars.new(self) + end + + # Returns true if the string has UTF-8 semantics (a String used for purely byte resources is unlikely to have + # them), returns false otherwise. + def is_utf8? + ActiveSupport::Multibyte::Handlers::UTF8Handler.consumes?(self) + end + end + end + end +end \ No newline at end of file diff --git a/activesupport/lib/active_support/multibyte.rb b/activesupport/lib/active_support/multibyte.rb new file mode 100644 index 0000000000..5e518b65f0 --- /dev/null +++ b/activesupport/lib/active_support/multibyte.rb @@ -0,0 +1,7 @@ +module ActiveSupport::Multibyte + DEFAULT_NORMALIZATION_FORM = :kc + NORMALIZATIONS_FORMS = [:c, :kc, :d, :kd] + UNICODE_VERSION = '5.0.0' +end + +require 'active_support/multibyte/chars' \ No newline at end of file diff --git a/activesupport/lib/active_support/multibyte/chars.rb b/activesupport/lib/active_support/multibyte/chars.rb new file mode 100644 index 0000000000..a9795de523 --- /dev/null +++ b/activesupport/lib/active_support/multibyte/chars.rb @@ -0,0 +1,129 @@ +require 'active_support/multibyte/handlers/utf8_handler' +require 'active_support/multibyte/handlers/passthru_handler' + +# Encapsulates all the functionality related to the Chars proxy. +module ActiveSupport::Multibyte + # Chars enables you to work transparently with multibyte encodings in the Ruby String class without having extensive + # knowledge about the encoding. A Chars object accepts a string upon initialization and proxies String methods in an + # encoding safe manner. All the normal String methods are also implemented on the proxy. + # + # String methods are proxied through the Chars object, and can be accessed through the +chars+ method. Methods + # which would normally return a String object now return a Chars object so methods can be chained. + # + # "The Perfect String ".chars.downcase.strip.normalize #=> "the perfect string" + # + # Chars objects are perfectly interchangeable with String objects as long as no explicit class checks are made. + # If certain methods do explicitly check the class, call +to_s+ before you pass chars objects to them. + # + # bad.explicit_checking_method "T".chars.downcase.to_s + # + # The actual operations on the string are delegated to handlers. Theoretically handlers can be implemented for + # any encoding, but the default handler handles UTF-8. This handler is set during initialization, if you want to + # use you own handler, you can set it on the Chars class. Look at the UTF8Handler source for an example how to + # implement your own handler. If you your own handler to work on anything but UTF-8 you probably also + # want to override Chars#handler. + # + # ActiveSupport::Multibyte::Chars.handler = MyHandler + # + # Note that a few methods are defined on Chars instead of the handler because they are defined on Object or Kernel + # and method_missing can't catch them. + class Chars + + attr_reader :string # The contained string + alias_method :to_s, :string + + include Comparable + + # The magic method to make String and Chars comparable + def to_str + # Using any other ways of overriding the String itself will lead you all the way from infinite loops to + # core dumps. Don't go there. + @string + end + + # Create a new Chars instance. + def initialize(str) + @string = (str.string rescue str) + end + + # Returns -1, 0 or +1 depending on whether the Chars object is to be sorted before, equal or after the + # object on the right side of the operation. It accepts any object that implements +to_s+. See String.<=> + # for more details. + def <=>(other); @string <=> other.to_s; end + + # Works just like String#split, with the exception that the items in the resulting list are Chars + # instances instead of String. This makes chaining methods easier. + def split(*args) + @string.split(*args).map { |i| i.chars } + end + + # Gsub works exactly the same as gsub on a normal string. + def gsub(*a, &b); @string.gsub(*a, &b).chars; end + + # Like String.=~ only it returns the character offset (in codepoints) instead of the byte offset. + def =~(other) + handler.translate_offset(@string, @string =~ other) + end + + # Try to forward all undefined methods to the handler, when a method is not defined on the handler, send it to + # the contained string. Method_missing is also responsible for making the bang! methods destructive. + def method_missing(m, *a, &b) + begin + # Simulate methods with a ! at the end because we can't touch the enclosed string from the handlers. + if m.to_s =~ /^(.*)\!$/ + result = handler.send($1, @string, *a, &b) + if result == @string + result = nil + else + @string.replace result + end + else + result = handler.send(m, @string, *a, &b) + end + rescue NoMethodError + result = @string.send(m, *a, &b) + rescue Handlers::EncodingError + @string.replace handler.tidy_bytes(@string) + retry + end + + if result.kind_of?(String) + result.chars + else + result + end + end + + # Set the handler class for the Char objects. + def self.handler=(klass) + @@handler = klass + end + + # Returns the proper handler for the contained string depending on $KCODE and the encoding of the string. This + # method is used internally to always redirect messages to the proper classes depending on the context. + def handler + if utf8_pragma? + @@handler + else + ActiveSupport::Multibyte::Handlers::PassthruHandler + end + end + + private + + # +utf8_pragma+ checks if it can send this string to the handlers. It makes sure @string isn't nil and $KCODE is + # set to 'UTF8'. + def utf8_pragma? + !@string.nil? && ($KCODE == 'UTF8') + end + end +end + +# When we can load the utf8proc library, override normalization with the faster methods +begin + require 'utf8proc_native' + require 'active_support/multibyte/handlers/utf8_handler_proc' + ActiveSupport::Multibyte::Chars.handler = ActiveSupport::Multibyte::Handlers::UTF8HandlerProc +rescue LoadError + ActiveSupport::Multibyte::Chars.handler = ActiveSupport::Multibyte::Handlers::UTF8Handler +end \ No newline at end of file diff --git a/activesupport/lib/active_support/multibyte/generators/generate_tables.rb b/activesupport/lib/active_support/multibyte/generators/generate_tables.rb new file mode 100644 index 0000000000..4045b94282 --- /dev/null +++ b/activesupport/lib/active_support/multibyte/generators/generate_tables.rb @@ -0,0 +1,141 @@ +#!/usr/bin/env ruby +begin + require File.dirname(__FILE__) + '/../../../active_support' +rescue IOError +end +require 'open-uri' +require 'tmpdir' + +module ActiveSupport::Multibyte::Handlers #:nodoc: + class UnicodeDatabase #:nodoc: + def self.load + [Hash.new(Codepoint.new),[],{},{}] + end + end + + class UnicodeTableGenerator #:nodoc: + BASE_URI = "http://www.unicode.org/Public/#{ActiveSupport::Multibyte::UNICODE_VERSION}/ucd/" + SOURCES = { + :codepoints => BASE_URI + 'UnicodeData.txt', + :composition_exclusion => BASE_URI + 'CompositionExclusions.txt', + :grapheme_break_property => BASE_URI + 'auxiliary/GraphemeBreakProperty.txt' + } + + def initialize + @ucd = UnicodeDatabase.new + + default = Codepoint.new + default.combining_class = 0 + default.uppercase_mapping = 0 + default.lowercase_mapping = 0 + @ucd.codepoints = Hash.new(default) + + @ucd.composition_exclusion = [] + @ucd.composition_map = {} + @ucd.boundary = {} + end + + def parse_codepoints(line) + codepoint = Codepoint.new + raise "Could not parse input." unless line =~ /^ + ([0-9A-F]+); # code + ([^;]+); # name + ([A-Z]+); # general category + ([0-9]+); # canonical combining class + ([A-Z]+); # bidi class + (<([A-Z]*)>)? # decomposition type + ((\ ?[0-9A-F]+)*); # decompomposition mapping + ([0-9]*); # decimal digit + ([0-9]*); # digit + ([^;]*); # numeric + ([YN]*); # bidi mirrored + ([^;]*); # unicode 1.0 name + ([^;]*); # iso comment + ([0-9A-F]*); # simple uppercase mapping + ([0-9A-F]*); # simple lowercase mapping + ([0-9A-F]*)$/ix # simple titlecase mapping + codepoint.code = $1.hex + #codepoint.name = $2 + #codepoint.category = $3 + codepoint.combining_class = Integer($4) + #codepoint.bidi_class = $5 + codepoint.decomp_type = $7 + codepoint.decomp_mapping = ($8=='') ? nil : $8.split.collect { |element| element.hex } + #codepoint.bidi_mirrored = ($13=='Y') ? true : false + codepoint.uppercase_mapping = ($16=='') ? 0 : $16.hex + codepoint.lowercase_mapping = ($17=='') ? 0 : $17.hex + #codepoint.titlecase_mapping = ($18=='') ? nil : $18.hex + @ucd.codepoints[codepoint.code] = codepoint + end + + def parse_grapheme_break_property(line) + if line =~ /^([0-9A-F\.]+)\s*;\s*([\w]+)\s*#/ + type = $2.downcase.intern + @ucd.boundary[type] ||= [] + if $1.include? '..' + parts = $1.split '..' + @ucd.boundary[type] << (parts[0].hex..parts[1].hex) + else + @ucd.boundary[type] << $1.hex + end + end + end + + def parse_composition_exclusion(line) + if line =~ /^([0-9A-F]+)/i + @ucd.composition_exclusion << $1.hex + end + end + + def create_composition_map + @ucd.codepoints.each do |_, cp| + if !cp.nil? and cp.combining_class == 0 and cp.decomp_type.nil? and !cp.decomp_mapping.nil? and cp.decomp_mapping.length == 2 and @ucd[cp.decomp_mapping[0]].combining_class == 0 and !@ucd.composition_exclusion.include?(cp.code) + @ucd.composition_map[cp.decomp_mapping[0]] ||= {} + @ucd.composition_map[cp.decomp_mapping[0]][cp.decomp_mapping[1]] = cp.code + end + end + end + + def normalize_boundary_map + @ucd.boundary.each do |k,v| + if [:lf, :cr].include? k + @ucd.boundary[k] = v[0] + end + end + end + + def parse + SOURCES.each do |type, url| + filename = File.join(Dir.tmpdir, "#{url.split('/').last}") + unless File.exist?(filename) + $stderr.puts "Downloading #{url.split('/').last}" + File.open(filename, 'wb') do |target| + open(url) do |source| + source.each_line { |line| target.write line } + end + end + end + File.open(filename) do |file| + file.each_line { |line| send "parse_#{type}".intern, line } + end + end + create_composition_map + normalize_boundary_map + end + + def dump_to(filename) + File.open(filename, 'wb') do |f| + f.write Marshal.dump([@ucd.codepoints, @ucd.composition_exclusion, @ucd.composition_map, @ucd.boundary]) + end + end + end +end + +if __FILE__ == $0 + filename = ActiveSupport::Multibyte::Handlers::UnicodeDatabase.filename + generator = ActiveSupport::Multibyte::Handlers::UnicodeTableGenerator.new + generator.parse + print "Writing to: #{filename}" + generator.dump_to filename + puts " (#{File.size(filename)} bytes)" +end diff --git a/activesupport/lib/active_support/multibyte/handlers/passthru_handler.rb b/activesupport/lib/active_support/multibyte/handlers/passthru_handler.rb new file mode 100644 index 0000000000..b28e345585 --- /dev/null +++ b/activesupport/lib/active_support/multibyte/handlers/passthru_handler.rb @@ -0,0 +1,9 @@ +# Chars uses this handler when $KCODE is not set to 'UTF8'. Because this handler doesn't define any methods all call +# will be forwarded to String. +class ActiveSupport::Multibyte::Handlers::PassthruHandler + + # Return the original byteoffset + def self.translate_offset(string, byte_offset) #:nodoc: + byte_offset + end +end \ No newline at end of file diff --git a/activesupport/lib/active_support/multibyte/handlers/utf8_handler.rb b/activesupport/lib/active_support/multibyte/handlers/utf8_handler.rb new file mode 100644 index 0000000000..0928209d85 --- /dev/null +++ b/activesupport/lib/active_support/multibyte/handlers/utf8_handler.rb @@ -0,0 +1,447 @@ +# Contains all the handlers and helper classes +module ActiveSupport::Multibyte::Handlers + class EncodingError < ArgumentError; end + + class Codepoint #:nodoc: + attr_accessor :code, :combining_class, :decomp_type, :decomp_mapping, :uppercase_mapping, :lowercase_mapping + end + + class UnicodeDatabase #:nodoc: + attr_accessor :codepoints, :composition_exclusion, :composition_map, :boundary + + # Creates a new UnicodeDatabase instance and loads the database. + def initialize + begin + @codepoints, @composition_exclusion, @composition_map, @boundary = self.class.load + rescue Exception => e + raise IOError.new("Couldn't load the unicode tables for UTF8Handler (#{e.message}), handler is unusable") + end + @codepoints ||= Hash.new(Codepoint.new) + @composition_exclusion ||= [] + @composition_map ||= {} + @boundary ||= {} + + # Redefine the === method so we can write shorter rules for grapheme cluster breaks + @boundary.each do |k,_| + @boundary[k].instance_eval do + def ===(other) + detect { |i| i === other } ? true : false + end + end if @boundary[k].kind_of?(Array) + end + end + + # Shortcut to ucd.codepoints[] + def [](index); @codepoints[index]; end + + # Returns the directory in which the data files are stored + def self.dirname + File.dirname(__FILE__) + '/../../values/' + end + + # Returns the filename for the data file for this version + def self.filename + File.expand_path File.join(dirname, "unicode_tables-#{VERSION}.dat") + end + + # Loads the unicode database and returns all the internal objects of UnicodeDatabase + def self.load + begin + return load_file(filename) + rescue Exception + # If we can't load our own version, try the rest + Dir["#{dirname}/*.dat"].sort.each do |dat| + begin + return load_file(dat) + rescue Exception + end + end + end + raise IOError.new("Can't load a marshal file for your version of Ruby") + end + + def self.load_file(filename) + File.open(self.filename, 'rb') { |f| Marshal.load f.read } + end + end + + # UTF8Handler implements Unicode aware operations for strings, these operations will be used by the Chars + # proxy when $KCODE is set to 'UTF8'. + class UTF8Handler + # Hangul character boundaries and properties + HANGUL_SBASE = 0xAC00 + HANGUL_LBASE = 0x1100 + HANGUL_VBASE = 0x1161 + HANGUL_TBASE = 0x11A7 + HANGUL_LCOUNT = 19 + HANGUL_VCOUNT = 21 + HANGUL_TCOUNT = 28 + HANGUL_NCOUNT = HANGUL_VCOUNT * HANGUL_TCOUNT + HANGUL_SCOUNT = 11172 + HANGUL_SLAST = HANGUL_SBASE + HANGUL_SCOUNT + HANGUL_JAMO_FIRST = 0x1100 + HANGUL_JAMO_LAST = 0x11FF + + # All the unicode whitespace + UNICODE_WHITESPACE = [ + (0x0009..0x000D).to_a, # White_Space # Cc [5] .. + 0x0020, # White_Space # Zs SPACE + 0x0085, # White_Space # Cc + 0x00A0, # White_Space # Zs NO-BREAK SPACE + 0x1680, # White_Space # Zs OGHAM SPACE MARK + 0x180E, # White_Space # Zs MONGOLIAN VOWEL SEPARATOR + (0x2000..0x200A).to_a, # White_Space # Zs [11] EN QUAD..HAIR SPACE + 0x2028, # White_Space # Zl LINE SEPARATOR + 0x2029, # White_Space # Zp PARAGRAPH SEPARATOR + 0x202F, # White_Space # Zs NARROW NO-BREAK SPACE + 0x205F, # White_Space # Zs MEDIUM MATHEMATICAL SPACE + 0x3000, # White_Space # Zs IDEOGRAPHIC SPACE + ].flatten.freeze + + # BOM (byte order mark) can also be seen as whitespace, it's a non-rendering character used to distinguish + # between little and big endian. This is not an issue in utf-8, so it must be ignored. + UNICODE_LEADERS_AND_TRAILERS = UNICODE_WHITESPACE + [65279] # ZERO-WIDTH NO-BREAK SPACE aka BOM + + # Borrowed from the Kconv library by Shinji KONO - (also as seen on the W3C site) + UTF8_PAT = /\A(?: + [\x00-\x7f] | + [\xc2-\xdf] [\x80-\xbf] | + \xe0 [\xa0-\xbf] [\x80-\xbf] | + [\xe1-\xef] [\x80-\xbf] [\x80-\xbf] | + \xf0 [\x90-\xbf] [\x80-\xbf] [\x80-\xbf] | + [\xf1-\xf3] [\x80-\xbf] [\x80-\xbf] [\x80-\xbf] | + \xf4 [\x80-\x8f] [\x80-\xbf] [\x80-\xbf] + )*\z/xn + + # Returns a regular expression pattern that matches the passed Unicode codepoints + def self.codepoints_to_pattern(array_of_codepoints) #:nodoc: + array_of_codepoints.collect{ |e| [e].pack 'U*' }.join('|') + end + UNICODE_TRAILERS_PAT = /(#{codepoints_to_pattern(UNICODE_LEADERS_AND_TRAILERS)})+\Z/ + UNICODE_LEADERS_PAT = /\A(#{codepoints_to_pattern(UNICODE_LEADERS_AND_TRAILERS)})+/ + + class << self + + # /// + # /// BEGIN String method overrides + # /// + + # Inserts the passed string at specified codepoint offsets + def insert(str, offset, fragment) + str.replace( + u_unpack(str).insert( + offset, + u_unpack(fragment) + ).flatten.pack('U*') + ) + end + + # Returns the position of the passed argument in the string, counting in codepoints + def index(str, *args) + bidx = str.index(*args) + bidx ? (u_unpack(str.slice(0...bidx)).size) : nil + end + + # Does Unicode-aware rstrip + def rstrip(str) + str.gsub(UNICODE_TRAILERS_PAT, '') + end + + # Does Unicode-aware lstrip + def lstrip(str) + str.gsub(UNICODE_LEADERS_PAT, '') + end + + # Removed leading and trailing whitespace + def strip(str) + str.gsub(UNICODE_LEADERS_PAT, '').gsub(UNICODE_TRAILERS_PAT, '') + end + + # Returns the number of codepoints in the string + def size(str) + u_unpack(str).size + end + alias_method :length, :size + + # Reverses codepoints in the string. + def reverse(str) + u_unpack(str).reverse.pack('U*') + end + + # Implements Unicode-aware slice with codepoints. Slicing on one point returns the codepoints for that + # character. + def slice(str, *args) + if (args.size == 2 && args.first.is_a?(Range)) + raise TypeError, 'cannot convert Range into Integer' # Do as if we were native + elsif args[0].kind_of? Range + cps = u_unpack(str).slice(*args) + cps.nil? ? nil : cps.pack('U*') + elsif args[0].kind_of? Numeric + u_unpack(str)[args[0]] + else + str.slice(*args) + end + end + alias_method :[], :slice + + # Convert characters in the string to uppercase + def upcase(str); to_case :uppercase_mapping, str; end + + # Convert characters in the string to lowercase + def downcase(str); to_case :lowercase_mapping, str; end + + # Returns a copy of +str+ with the first character converted to uppercase and the remainder to lowercase + def capitalize(str) + upcase(slice(str, 0..0)) + downcase(slice(str, 1..-1) || '') + end + + # /// + # /// Extra String methods for unicode operations + # /// + + # Returns the KC normalization of the string by default. NFKC is considered the best normalization form for + # passing strings to databases and validations. + # + # * str: The string to perform normalization on. + # * form: The form you want to normalize in. Should be one of the following: :c, :kc, :d or :kd. + def normalize(str, form=ActiveSupport::Multibyte::DEFAULT_NORMALIZATION_FORM) + # See http://www.unicode.org/reports/tr15, Table 1 + codepoints = u_unpack(str) + case form + when :d + reorder_characters(decompose_codepoints(:canonical, codepoints)) + when :c + compose_codepoints reorder_characters(decompose_codepoints(:canonical, codepoints)) + when :kd + reorder_characters(decompose_codepoints(:compatability, codepoints)) + when :kc + compose_codepoints reorder_characters(decompose_codepoints(:compatability, codepoints)) + else + raise ArgumentError, "#{form} is not a valid normalization variant", caller + end.pack('U*') + end + + # Perform decomposition on the characters in the string + def decompose(str) + decompose_codepoints(:canonical, u_unpack(str)).pack('U*') + end + + # Perform composition on the characters in the string + def compose(str) + compose_codepoints u_unpack(str).pack('U*') + end + + # /// + # /// BEGIN Helper methods for unicode operation + # /// + + # Used to translate an offset from bytes to characters, for instance one received from a regular expression match + def translate_offset(str, byte_offset) + return 0 if str == '' + return nil if byte_offset.nil? + chunk = str[0..byte_offset] + begin + begin + chunk.unpack('U*').length - 1 + rescue ArgumentError => e + chunk = str[0..(byte_offset+=1)] + # Stop retrying at the end of the string + raise e unless byte_offset < chunk.length + # We damaged a character, retry + retry + end + # Catch the ArgumentError so we can throw our own + rescue ArgumentError + raise EncodingError.new('malformed UTF-8 character') + end + end + + # Checks if the string is valid UTF8. + def consumes?(str) + # Unpack is a little bit faster than regular expressions + begin + str.unpack('U*') + true + rescue ArgumentError + false + end + end + + # Returns the number of grapheme clusters in the string. This method is very likely to be moved or renamed + # in future versions. + def g_length(str) + g_unpack(str).length + end + + # Strips all the non-utf-8 bytes from the string resulting in a valid utf-8 string + def tidy_bytes(str) + str.split(//u).reject { |c| !UTF8_PAT.match(c) }.join + end + + protected + + # Detect whether the codepoint is in a certain character class. Primarily used by the + # grapheme cluster support. + def in_char_class?(codepoint, classes) + classes.detect { |c| UCD.boundary[c] === codepoint } ? true : false + end + + # Unpack the string at codepoints boundaries + def u_unpack(str) + begin + str.unpack 'U*' + rescue ArgumentError + raise EncodingError.new('malformed UTF-8 character') + end + end + + # Unpack the string at grapheme boundaries instead of codepoint boundaries + def g_unpack(str) + codepoints = u_unpack(str) + unpacked = [] + pos = 0 + marker = 0 + eoc = codepoints.length + while(pos < eoc) + pos += 1 + previous = codepoints[pos-1] + current = codepoints[pos] + if ( + # CR X LF + one = ( previous == UCD.boundary[:cr] and current == UCD.boundary[:lf] ) or + # L X (L|V|LV|LVT) + two = ( UCD.boundary[:l] === previous and in_char_class?(current, [:l,:v,:lv,:lvt]) ) or + # (LV|V) X (V|T) + three = ( in_char_class?(previous, [:lv,:v]) and in_char_class?(current, [:v,:t]) ) or + # (LVT|T) X (T) + four = ( in_char_class?(previous, [:lvt,:t]) and UCD.boundary[:t] === current ) or + # X Extend + five = (UCD.boundary[:extend] === current) + ) + else + unpacked << codepoints[marker..pos-1] + marker = pos + end + end + unpacked + end + + # Reverse operation of g_unpack + def g_pack(unpacked) + unpacked.flatten + end + + # Convert characters to a different case + def to_case(way, str) + u_unpack(str).map do |codepoint| + cp = UCD[codepoint] + unless cp.nil? + ncp = cp.send(way) + ncp > 0 ? ncp : codepoint + else + codepoint + end + end.pack('U*') + end + + # Re-order codepoints so the string becomes canonical + def reorder_characters(codepoints) + length = codepoints.length- 1 + pos = 0 + while pos < length do + cp1, cp2 = UCD[codepoints[pos]], UCD[codepoints[pos+1]] + if (cp1.combining_class > cp2.combining_class) && (cp2.combining_class > 0) + codepoints[pos..pos+1] = cp2.code, cp1.code + pos += (pos > 0 ? -1 : 1) + else + pos += 1 + end + end + codepoints + end + + # Decompose composed characters to the decomposed form + def decompose_codepoints(type, codepoints) + codepoints.inject([]) do |decomposed, cp| + # if it's a hangul syllable starter character + if HANGUL_SBASE <= cp and cp < HANGUL_SLAST + sindex = cp - HANGUL_SBASE + ncp = [] # new codepoints + ncp << HANGUL_LBASE + sindex / HANGUL_NCOUNT + ncp << HANGUL_VBASE + (sindex % HANGUL_NCOUNT) / HANGUL_TCOUNT + tindex = sindex % HANGUL_TCOUNT + ncp << (HANGUL_TBASE + tindex) unless tindex == 0 + decomposed.concat ncp + # if the codepoint is decomposable in with the current decomposition type + elsif (ncp = UCD[cp].decomp_mapping) and (!UCD[cp].decomp_type || type == :compatability) + decomposed.concat decompose_codepoints(type, ncp.dup) + else + decomposed << cp + end + end + end + + # Compose decomposed characters to the composed form + def compose_codepoints(codepoints) + pos = 0 + eoa = codepoints.length - 1 + starter_pos = 0 + starter_char = codepoints[0] + previous_combining_class = -1 + while pos < eoa + pos += 1 + lindex = starter_char - HANGUL_LBASE + # -- Hangul + if 0 <= lindex and lindex < HANGUL_LCOUNT + vindex = codepoints[starter_pos+1] - HANGUL_VBASE rescue vindex = -1 + if 0 <= vindex and vindex < HANGUL_VCOUNT + tindex = codepoints[starter_pos+2] - HANGUL_TBASE rescue tindex = -1 + if 0 <= tindex and tindex < HANGUL_TCOUNT + j = starter_pos + 2 + eoa -= 2 + else + tindex = 0 + j = starter_pos + 1 + eoa -= 1 + end + codepoints[starter_pos..j] = (lindex * HANGUL_VCOUNT + vindex) * HANGUL_TCOUNT + tindex + HANGUL_SBASE + end + starter_pos += 1 + starter_char = codepoints[starter_pos] + # -- Other characters + else + current_char = codepoints[pos] + current = UCD[current_char] + if current.combining_class > previous_combining_class + if ref = UCD.composition_map[starter_char] + composition = ref[current_char] + else + composition = nil + end + unless composition.nil? + codepoints[starter_pos] = composition + starter_char = composition + codepoints.delete_at pos + eoa -= 1 + pos -= 1 + previous_combining_class = -1 + else + previous_combining_class = current.combining_class + end + else + previous_combining_class = current.combining_class + end + if current.combining_class == 0 + starter_pos = pos + starter_char = codepoints[pos] + end + end + end + codepoints + end + + # UniCode Database + UCD = UnicodeDatabase.new + end + end +end \ No newline at end of file diff --git a/activesupport/lib/active_support/multibyte/handlers/utf8_handler_proc.rb b/activesupport/lib/active_support/multibyte/handlers/utf8_handler_proc.rb new file mode 100644 index 0000000000..f551af6cc7 --- /dev/null +++ b/activesupport/lib/active_support/multibyte/handlers/utf8_handler_proc.rb @@ -0,0 +1,44 @@ +# Methods in this handler call functions in the utf8proc ruby extension. These are significantly faster than the +# pure ruby versions. Chars automatically uses this handler when it can load the utf8proc extension. For +# documentation on handler methods see UTF8Handler. +class ActiveSupport::Multibyte::Handlers::UTF8HandlerProc < ActiveSupport::Multibyte::Handlers::UTF8Handler + + class << self + def normalize(str, form=ActiveSupport::Multibyte::DEFAULT_NORMALIZATION_FORM) #:nodoc: + codepoints = str.unpack('U*') + case form + when :d + utf8map(str, :stable) + when :c + utf8map(str, :stable, :compose) + when :kd + utf8map(str, :stable, :compat) + when :kc + utf8map(str, :stable, :compose, :compat) + else + raise ArgumentError, "#{form} is not a valid normalization variant", caller + end + end + + def decompose(str) #:nodoc: + utf8map(str, :stable) + end + + def downcase(str) #:nodoc:c + utf8map(str, :casefold) + end + + protected + + def utf8map(str, *option_array) #:nodoc: + options = 0 + option_array.each do |option| + flag = Utf8Proc::Options[option] + raise ArgumentError, "Unknown argument given to utf8map." unless + flag + options |= flag + end + return Utf8Proc::utf8map(str, options) + end + end +end diff --git a/activesupport/lib/active_support/values/unicode_tables-1.8.4.dat b/activesupport/lib/active_support/values/unicode_tables-1.8.4.dat new file mode 100644 index 0000000000..13a081388a Binary files /dev/null and b/activesupport/lib/active_support/values/unicode_tables-1.8.4.dat differ diff --git a/activesupport/lib/active_support/values/unicode_tables-1.8.5.dat b/activesupport/lib/active_support/values/unicode_tables-1.8.5.dat new file mode 100644 index 0000000000..7b96885f32 Binary files /dev/null and b/activesupport/lib/active_support/values/unicode_tables-1.8.5.dat differ -- cgit v1.2.3