3
module ActiveSupport #:nodoc:
4
module Multibyte #:nodoc:
5
# Chars enables you to work transparently with UTF-8 encoding in the Ruby String class without having extensive
6
# knowledge about the encoding. A Chars object accepts a string upon initialization and proxies String methods in an
7
# encoding safe manner. All the normal String methods are also implemented on the proxy.
9
# String methods are proxied through the Chars object, and can be accessed through the +mb_chars+ method. Methods
10
# which would normally return a String object now return a Chars object so methods can be chained.
12
# "The Perfect String ".mb_chars.downcase.strip.normalize #=> "the perfect string"
14
# Chars objects are perfectly interchangeable with String objects as long as no explicit class checks are made.
15
# If certain methods do explicitly check the class, call +to_s+ before you pass chars objects to them.
17
# bad.explicit_checking_method "T".mb_chars.downcase.to_s
19
# The default Chars implementation assumes that the encoding of the string is UTF-8, if you want to handle different
20
# encodings you can write your own multibyte string handler and configure it through
21
# ActiveSupport::Multibyte.proxy_class.
25
# @wrapped_string.size / 4
28
# def self.accepts?(string)
29
# string.length % 4 == 0
33
# ActiveSupport::Multibyte.proxy_class = CharsForUTF32
35
# Hangul character boundaries and properties
43
HANGUL_NCOUNT = HANGUL_VCOUNT * HANGUL_TCOUNT
45
HANGUL_SLAST = HANGUL_SBASE + HANGUL_SCOUNT
46
HANGUL_JAMO_FIRST = 0x1100
47
HANGUL_JAMO_LAST = 0x11FF
49
# All the unicode whitespace
50
UNICODE_WHITESPACE = [
51
(0x0009..0x000D).to_a, # White_Space # Cc [5] <control-0009>..<control-000D>
52
0x0020, # White_Space # Zs SPACE
53
0x0085, # White_Space # Cc <control-0085>
54
0x00A0, # White_Space # Zs NO-BREAK SPACE
55
0x1680, # White_Space # Zs OGHAM SPACE MARK
56
0x180E, # White_Space # Zs MONGOLIAN VOWEL SEPARATOR
57
(0x2000..0x200A).to_a, # White_Space # Zs [11] EN QUAD..HAIR SPACE
58
0x2028, # White_Space # Zl LINE SEPARATOR
59
0x2029, # White_Space # Zp PARAGRAPH SEPARATOR
60
0x202F, # White_Space # Zs NARROW NO-BREAK SPACE
61
0x205F, # White_Space # Zs MEDIUM MATHEMATICAL SPACE
62
0x3000, # White_Space # Zs IDEOGRAPHIC SPACE
65
# BOM (byte order mark) can also be seen as whitespace, it's a non-rendering character used to distinguish
66
# between little and big endian. This is not an issue in utf-8, so it must be ignored.
67
UNICODE_LEADERS_AND_TRAILERS = UNICODE_WHITESPACE + [65279] # ZERO-WIDTH NO-BREAK SPACE aka BOM
69
# Returns a regular expression pattern that matches the passed Unicode codepoints
70
def self.codepoints_to_pattern(array_of_codepoints) #:nodoc:
71
array_of_codepoints.collect{ |e| [e].pack 'U*' }.join('|')
73
UNICODE_TRAILERS_PAT = /(#{codepoints_to_pattern(UNICODE_LEADERS_AND_TRAILERS)})+\Z/
74
UNICODE_LEADERS_PAT = /\A(#{codepoints_to_pattern(UNICODE_LEADERS_AND_TRAILERS)})+/
76
UTF8_PAT = ActiveSupport::Multibyte::VALID_CHARACTER['UTF-8']
78
attr_reader :wrapped_string
79
alias to_s wrapped_string
80
alias to_str wrapped_string
82
if '1.9'.respond_to?(:force_encoding)
83
# Creates a new Chars instance by wrapping _string_.
84
def initialize(string)
85
@wrapped_string = string
86
@wrapped_string.force_encoding(Encoding::UTF_8) unless @wrapped_string.frozen?
89
def initialize(string) #:nodoc:
90
@wrapped_string = string
94
# Forward all undefined methods to the wrapped string.
95
def method_missing(method, *args, &block)
96
if method.to_s =~ /!$/
97
@wrapped_string.__send__(method, *args, &block)
100
result = @wrapped_string.__send__(method, *args, &block)
101
result.kind_of?(String) ? chars(result) : result
105
# Returns +true+ if _obj_ responds to the given method. Private methods are included in the search
106
# only if the optional second parameter evaluates to +true+.
107
def respond_to?(method, include_private=false)
108
super || @wrapped_string.respond_to?(method, include_private) || false
111
# Enable more predictable duck-typing on String-like classes. See Object#acts_like?.
112
def acts_like_string?
116
# Returns +true+ if the Chars class can and should act as a proxy for the string _string_. Returns
118
def self.wants?(string)
119
$KCODE == 'UTF8' && consumes?(string)
122
# Returns +true+ when the proxy class can handle the string. Returns +false+ otherwise.
123
def self.consumes?(string)
124
# Unpack is a little bit faster than regular expressions.
133
# Returns <tt>-1</tt>, <tt>0</tt> or <tt>+1</tt> depending on whether the Chars object is to be sorted before,
134
# equal or after the object on the right side of the operation. It accepts any object that implements +to_s+.
135
# See <tt>String#<=></tt> for more details.
138
# 'é'.mb_chars <=> 'ü'.mb_chars #=> -1
140
@wrapped_string <=> other.to_s
143
# Returns a new Chars object containing the _other_ object concatenated to the string.
146
# ('Café'.mb_chars + ' périferôl').to_s #=> "Café périferôl"
151
# Like <tt>String#=~</tt> only it returns the character offset (in codepoints) instead of the byte offset.
154
# 'Café périferôl'.mb_chars =~ /ô/ #=> 12
156
translate_offset(@wrapped_string =~ other)
159
# Works just like <tt>String#split</tt>, with the exception that the items in the resulting list are Chars
160
# instances instead of String. This makes chaining methods easier.
163
# 'Café périferôl'.mb_chars.split(/é/).map { |part| part.upcase.to_s } #=> ["CAF", " P", "RIFERÔL"]
165
@wrapped_string.split(*args).map { |i| i.mb_chars }
168
# Inserts the passed string at specified codepoint offsets.
171
# 'Café'.mb_chars.insert(4, ' périferôl').to_s #=> "Café périferôl"
172
def insert(offset, fragment)
173
unpacked = self.class.u_unpack(@wrapped_string)
174
unless offset > unpacked.length
175
@wrapped_string.replace(
176
self.class.u_unpack(@wrapped_string).insert(offset, *self.class.u_unpack(fragment)).pack('U*')
179
raise IndexError, "index #{offset} out of string"
184
# Returns +true+ if contained string contains _other_. Returns +false+ otherwise.
187
# 'Café'.mb_chars.include?('é') #=> true
189
# We have to redefine this method because Enumerable defines it.
190
@wrapped_string.include?(other)
193
# Returns the position _needle_ in the string, counting in codepoints. Returns +nil+ if _needle_ isn't found.
196
# 'Café périferôl'.mb_chars.index('ô') #=> 12
197
# 'Café périferôl'.mb_chars.index(/\w/u) #=> 0
198
def index(needle, offset=0)
199
wrapped_offset = self.first(offset).wrapped_string.length
200
index = @wrapped_string.index(needle, wrapped_offset)
201
index ? (self.class.u_unpack(@wrapped_string.slice(0...index)).size) : nil
204
# Returns the position _needle_ in the string, counting in
205
# codepoints, searching backward from _offset_ or the end of the
206
# string. Returns +nil+ if _needle_ isn't found.
209
# 'Café périferôl'.mb_chars.rindex('é') #=> 6
210
# 'Café périferôl'.mb_chars.rindex(/\w/u) #=> 13
211
def rindex(needle, offset=nil)
213
wrapped_offset = self.first(offset).wrapped_string.length
214
index = @wrapped_string.rindex(needle, wrapped_offset)
215
index ? (self.class.u_unpack(@wrapped_string.slice(0...index)).size) : nil
218
# Like <tt>String#[]=</tt>, except instead of byte offsets you specify character offsets.
223
# s.mb_chars[2] = "e" # Replace character with offset 2
228
# s.mb_chars[1, 2] = "ö" # Replace 2 characters at character offset 1
232
replace_by = args.pop
233
# Indexed replace with regular expressions already works
234
if args.first.is_a?(Regexp)
235
@wrapped_string[*args] = replace_by
237
result = self.class.u_unpack(@wrapped_string)
238
if args[0].is_a?(Fixnum)
239
raise IndexError, "index #{args[0]} out of string" if args[0] >= result.length
241
max = args[1].nil? ? min : (min + args[1] - 1)
242
range = Range.new(min, max)
243
replace_by = [replace_by].pack('U') if replace_by.is_a?(Fixnum)
244
elsif args.first.is_a?(Range)
245
raise RangeError, "#{args[0]} out of range" if args[0].min >= result.length
248
needle = args[0].to_s
250
max = min + self.class.u_unpack(needle).length - 1
251
range = Range.new(min, max)
253
result[range] = self.class.u_unpack(replace_by)
254
@wrapped_string.replace(result.pack('U*'))
258
# Works just like <tt>String#rjust</tt>, only integer specifies characters instead of bytes.
262
# "¾ cup".mb_chars.rjust(8).to_s
265
# "¾ cup".mb_chars.rjust(8, " ").to_s # Use non-breaking whitespace
267
def rjust(integer, padstr=' ')
268
justify(integer, :right, padstr)
271
# Works just like <tt>String#ljust</tt>, only integer specifies characters instead of bytes.
275
# "¾ cup".mb_chars.rjust(8).to_s
278
# "¾ cup".mb_chars.rjust(8, " ").to_s # Use non-breaking whitespace
280
def ljust(integer, padstr=' ')
281
justify(integer, :left, padstr)
284
# Works just like <tt>String#center</tt>, only integer specifies characters instead of bytes.
288
# "¾ cup".mb_chars.center(8).to_s
291
# "¾ cup".mb_chars.center(8, " ").to_s # Use non-breaking whitespace
293
def center(integer, padstr=' ')
294
justify(integer, :center, padstr)
297
# Strips entire range of Unicode whitespace from the right of the string.
299
chars(@wrapped_string.gsub(UNICODE_TRAILERS_PAT, ''))
302
# Strips entire range of Unicode whitespace from the left of the string.
304
chars(@wrapped_string.gsub(UNICODE_LEADERS_PAT, ''))
307
# Strips entire range of Unicode whitespace from the right and left of the string.
312
# Returns the number of codepoints in the string
314
self.class.u_unpack(@wrapped_string).size
316
alias_method :length, :size
318
# Reverses all characters in the string.
321
# 'Café'.mb_chars.reverse.to_s #=> 'éfaC'
323
chars(self.class.g_unpack(@wrapped_string).reverse.flatten.pack('U*'))
326
# Implements Unicode-aware slice with codepoints. Slicing on one point returns the codepoints for that
330
# 'こんにちは'.mb_chars.slice(2..3).to_s #=> "にち"
333
raise ArgumentError, "wrong number of arguments (#{args.size} for 1)" # Do as if we were native
334
elsif (args.size == 2 && !(args.first.is_a?(Numeric) || args.first.is_a?(Regexp)))
335
raise TypeError, "cannot convert #{args.first.class} into Integer" # Do as if we were native
336
elsif (args.size == 2 && !args[1].is_a?(Numeric))
337
raise TypeError, "cannot convert #{args[1].class} into Integer" # Do as if we were native
338
elsif args[0].kind_of? Range
339
cps = self.class.u_unpack(@wrapped_string).slice(*args)
340
result = cps.nil? ? nil : cps.pack('U*')
341
elsif args[0].kind_of? Regexp
342
result = @wrapped_string.slice(*args)
343
elsif args.size == 1 && args[0].kind_of?(Numeric)
344
character = self.class.u_unpack(@wrapped_string)[args[0]]
345
result = character.nil? ? nil : [character].pack('U')
347
result = self.class.u_unpack(@wrapped_string).slice(*args).pack('U*')
349
result.nil? ? nil : chars(result)
351
alias_method :[], :slice
353
# Like <tt>String#slice!</tt>, except instead of byte offsets you specify character offsets.
357
# s.mb_chars.slice!(2..3).to_s #=> "にち"
365
# Returns the codepoint of the first character in the string.
368
# 'こんにちは'.mb_chars.ord #=> 12371
370
self.class.u_unpack(@wrapped_string)[0]
373
# Convert characters in the string to uppercase.
376
# 'Laurent, òu sont les tests?'.mb_chars.upcase.to_s #=> "LAURENT, ÒU SONT LES TESTS?"
378
apply_mapping :uppercase_mapping
381
# Convert characters in the string to lowercase.
384
# 'VĚDA A VÝZKUM'.mb_chars.downcase.to_s #=> "věda a výzkum"
386
apply_mapping :lowercase_mapping
389
# Converts the first character to uppercase and the remainder to lowercase.
392
# 'über'.mb_chars.capitalize.to_s #=> "Über"
394
(slice(0) || chars('')).upcase + (slice(1..-1) || chars('')).downcase
397
# Returns the KC normalization of the string by default. NFKC is considered the best normalization form for
398
# passing strings to databases and validations.
400
# * <tt>str</tt> - The string to perform normalization on.
401
# * <tt>form</tt> - The form you want to normalize in. Should be one of the following:
402
# <tt>:c</tt>, <tt>:kc</tt>, <tt>:d</tt>, or <tt>:kd</tt>. Default is
403
# ActiveSupport::Multibyte.default_normalization_form
404
def normalize(form=ActiveSupport::Multibyte.default_normalization_form)
405
# See http://www.unicode.org/reports/tr15, Table 1
406
codepoints = self.class.u_unpack(@wrapped_string)
409
self.class.reorder_characters(self.class.decompose_codepoints(:canonical, codepoints))
411
self.class.compose_codepoints(self.class.reorder_characters(self.class.decompose_codepoints(:canonical, codepoints)))
413
self.class.reorder_characters(self.class.decompose_codepoints(:compatability, codepoints))
415
self.class.compose_codepoints(self.class.reorder_characters(self.class.decompose_codepoints(:compatability, codepoints)))
417
raise ArgumentError, "#{form} is not a valid normalization variant", caller
421
# Performs canonical decomposition on all the characters.
425
# 'é'.mb_chars.decompose.to_s.length #=> 3
427
chars(self.class.decompose_codepoints(:canonical, self.class.u_unpack(@wrapped_string)).pack('U*'))
430
# Performs composition on all the characters.
434
# 'é'.mb_chars.compose.to_s.length #=> 2
436
chars(self.class.compose_codepoints(self.class.u_unpack(@wrapped_string)).pack('U*'))
439
# Returns the number of grapheme clusters in the string.
442
# 'क्षि'.mb_chars.length #=> 4
443
# 'क्षि'.mb_chars.g_length #=> 3
445
self.class.g_unpack(@wrapped_string).length
448
# Replaces all ISO-8859-1 or CP1252 characters by their UTF-8 equivalent resulting in a valid UTF-8 string.
450
chars(self.class.tidy_bytes(@wrapped_string))
453
%w(lstrip rstrip strip reverse upcase downcase tidy_bytes capitalize).each do |method|
454
define_method("#{method}!") do |*args|
456
@wrapped_string = send(method, *args).to_s
458
@wrapped_string = send(method).to_s
466
# Unpack the string at codepoints boundaries. Raises an EncodingError when the encoding of the string isn't
470
# Chars.u_unpack('Café') #=> [67, 97, 102, 233]
475
raise EncodingError, 'malformed UTF-8 character'
479
# Detect whether the codepoint is in a certain character class. Returns +true+ when it's in the specified
480
# character class and +false+ otherwise. Valid character classes are: <tt>:cr</tt>, <tt>:lf</tt>, <tt>:l</tt>,
481
# <tt>:v</tt>, <tt>:lv</tt>, <tt>:lvt</tt> and <tt>:t</tt>.
483
# Primarily used by the grapheme cluster support.
484
def in_char_class?(codepoint, classes)
485
classes.detect { |c| UCD.boundary[c] === codepoint } ? true : false
488
# Unpack the string at grapheme boundaries. Returns a list of character lists.
491
# Chars.g_unpack('क्षि') #=> [[2325, 2381], [2359], [2367]]
492
# Chars.g_unpack('Café') #=> [[67], [97], [102], [233]]
494
codepoints = u_unpack(string)
498
eoc = codepoints.length
501
previous = codepoints[pos-1]
502
current = codepoints[pos]
505
one = ( previous == UCD.boundary[:cr] and current == UCD.boundary[:lf] ) or
507
two = ( UCD.boundary[:l] === previous and in_char_class?(current, [:l,:v,:lv,:lvt]) ) or
509
three = ( in_char_class?(previous, [:lv,:v]) and in_char_class?(current, [:v,:t]) ) or
511
four = ( in_char_class?(previous, [:lvt,:t]) and UCD.boundary[:t] === current ) or
513
five = (UCD.boundary[:extend] === current)
516
unpacked << codepoints[marker..pos-1]
523
# Reverse operation of g_unpack.
526
# Chars.g_pack(Chars.g_unpack('क्षि')) #=> 'क्षि'
528
(unpacked.flatten).pack('U*')
531
def padding(padsize, padstr=' ') #:nodoc:
533
new(padstr * ((padsize / u_unpack(padstr).size) + 1)).slice(0, padsize)
539
# Re-order codepoints so the string becomes canonical.
540
def reorder_characters(codepoints)
541
length = codepoints.length- 1
543
while pos < length do
544
cp1, cp2 = UCD.codepoints[codepoints[pos]], UCD.codepoints[codepoints[pos+1]]
545
if (cp1.combining_class > cp2.combining_class) && (cp2.combining_class > 0)
546
codepoints[pos..pos+1] = cp2.code, cp1.code
547
pos += (pos > 0 ? -1 : 1)
555
# Decompose composed characters to the decomposed form.
556
def decompose_codepoints(type, codepoints)
557
codepoints.inject([]) do |decomposed, cp|
558
# if it's a hangul syllable starter character
559
if HANGUL_SBASE <= cp and cp < HANGUL_SLAST
560
sindex = cp - HANGUL_SBASE
561
ncp = [] # new codepoints
562
ncp << HANGUL_LBASE + sindex / HANGUL_NCOUNT
563
ncp << HANGUL_VBASE + (sindex % HANGUL_NCOUNT) / HANGUL_TCOUNT
564
tindex = sindex % HANGUL_TCOUNT
565
ncp << (HANGUL_TBASE + tindex) unless tindex == 0
566
decomposed.concat ncp
567
# if the codepoint is decomposable in with the current decomposition type
568
elsif (ncp = UCD.codepoints[cp].decomp_mapping) and (!UCD.codepoints[cp].decomp_type || type == :compatability)
569
decomposed.concat decompose_codepoints(type, ncp.dup)
576
# Compose decomposed characters to the composed form.
577
def compose_codepoints(codepoints)
579
eoa = codepoints.length - 1
581
starter_char = codepoints[0]
582
previous_combining_class = -1
585
lindex = starter_char - HANGUL_LBASE
587
if 0 <= lindex and lindex < HANGUL_LCOUNT
588
vindex = codepoints[starter_pos+1] - HANGUL_VBASE rescue vindex = -1
589
if 0 <= vindex and vindex < HANGUL_VCOUNT
590
tindex = codepoints[starter_pos+2] - HANGUL_TBASE rescue tindex = -1
591
if 0 <= tindex and tindex < HANGUL_TCOUNT
599
codepoints[starter_pos..j] = (lindex * HANGUL_VCOUNT + vindex) * HANGUL_TCOUNT + tindex + HANGUL_SBASE
602
starter_char = codepoints[starter_pos]
603
# -- Other characters
605
current_char = codepoints[pos]
606
current = UCD.codepoints[current_char]
607
if current.combining_class > previous_combining_class
608
if ref = UCD.composition_map[starter_char]
609
composition = ref[current_char]
613
unless composition.nil?
614
codepoints[starter_pos] = composition
615
starter_char = composition
616
codepoints.delete_at pos
619
previous_combining_class = -1
621
previous_combining_class = current.combining_class
624
previous_combining_class = current.combining_class
626
if current.combining_class == 0
628
starter_char = codepoints[pos]
635
# Replaces all ISO-8859-1 or CP1252 characters by their UTF-8 equivalent resulting in a valid UTF-8 string.
636
def tidy_bytes(string)
637
string.split(//u).map do |c|
638
c.force_encoding(Encoding::ASCII) if c.respond_to?(:force_encoding)
640
if !ActiveSupport::Multibyte::VALID_CHARACTER['UTF-8'].match(c)
643
n < 160 ? [UCD.cp1252[n] || n].pack('U') :
644
n < 192 ? "\xC2" + n.chr : "\xC3" + (n-64).chr
654
def translate_offset(byte_offset) #:nodoc:
655
return nil if byte_offset.nil?
656
return 0 if @wrapped_string == ''
657
chunk = @wrapped_string[0..byte_offset]
660
chunk.unpack('U*').length - 1
661
rescue ArgumentError => e
662
chunk = @wrapped_string[0..(byte_offset+=1)]
663
# Stop retrying at the end of the string
664
raise e unless byte_offset < chunk.length
665
# We damaged a character, retry
668
# Catch the ArgumentError so we can throw our own
670
raise EncodingError, 'malformed UTF-8 character'
674
def justify(integer, way, padstr=' ') #:nodoc:
675
raise ArgumentError, "zero width padding" if padstr.length == 0
676
padsize = integer - size
677
padsize = padsize > 0 ? padsize : 0
680
result = @wrapped_string.dup.insert(0, self.class.padding(padsize, padstr))
682
result = @wrapped_string.dup.insert(-1, self.class.padding(padsize, padstr))
684
lpad = self.class.padding((padsize / 2.0).floor, padstr)
685
rpad = self.class.padding((padsize / 2.0).ceil, padstr)
686
result = @wrapped_string.dup.insert(0, lpad).insert(-1, rpad)
691
def apply_mapping(mapping) #:nodoc:
692
chars(self.class.u_unpack(@wrapped_string).map do |codepoint|
693
cp = UCD.codepoints[codepoint]
694
if cp and (ncp = cp.send(mapping)) and ncp > 0
702
def chars(string) #:nodoc:
703
self.class.new(string)