?  search:
summary | log | graphiclog1 | graphiclog2 | commit | commitdiff | tree | refs | edit | fork
blame | history | raw | HEAD
Merge "Whitelist the <wbr> element."
[mediawiki.git] / includes / normal / UtfNormal.php
blob5a091afc96e162dbcb65d0af936cb6289392051b
1 <?php
2 /**
3 * Unicode normalization routines
4 *
5 * Copyright © 2004 Brion Vibber <brion@pobox.com>
6 * http://www.mediawiki.org/
7 *
8 * This program is free software; you can redistribute it and/or modify
9 * it under the terms of the GNU General Public License as published by
10 * the Free Software Foundation; either version 2 of the License, or
11 * (at your option) any later version.
12 *
13 * This program is distributed in the hope that it will be useful,
14 * but WITHOUT ANY WARRANTY; without even the implied warranty of
15 * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
16 * GNU General Public License for more details.
17 *
18 * You should have received a copy of the GNU General Public License along
19 * with this program; if not, write to the Free Software Foundation, Inc.,
20 * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
21 * http://www.gnu.org/copyleft/gpl.html
22 *
23 * @file
24 * @ingroup UtfNormal
25 */
26
27 /**
28 * @defgroup UtfNormal UtfNormal
29 */
30
31 define( 'NORMALIZE_ICU', function_exists( 'utf8_normalize' ) );
32 define( 'NORMALIZE_INTL', function_exists( 'normalizer_normalize' ) );
33
34 /**
35 * Unicode normalization routines for working with UTF-8 strings.
36 * Currently assumes that input strings are valid UTF-8!
37 *
38 * Not as fast as I'd like, but should be usable for most purposes.
39 * UtfNormal::toNFC() will bail early if given ASCII text or text
40 * it can quickly determine is already normalized.
41 *
42 * All functions can be called static.
43 *
44 * See description of forms at http://www.unicode.org/reports/tr15/
45 *
46 * @ingroup UtfNormal
47 */
48 class UtfNormal {
49 /**
50 * For using the ICU wrapper
51 */
52 const UNORM_NONE = 1;
53 const UNORM_NFD = 2;
54 const UNORM_NFKD = 3;
55 const UNORM_NFC = 4;
56 const UNORM_NFKC = 5;
57 const UNORM_FCD = 6;
58 const UNORM_DEFAULT = self::UNORM_NFC;
59
60 static $utfCombiningClass = null;
61 static $utfCanonicalComp = null;
62 static $utfCanonicalDecomp = null;
63
64 # Load compatibility decompositions on demand if they are needed.
65 static $utfCompatibilityDecomp = null;
66
67 static $utfCheckNFC;
68
69 /**
70 * The ultimate convenience function! Clean up invalid UTF-8 sequences,
71 * and convert to normal form C, canonical composition.
72 *
73 * Fast return for pure ASCII strings; some lesser optimizations for
74 * strings containing only known-good characters. Not as fast as toNFC().
75 *
76 * @param string $string a UTF-8 string
77 * @return string a clean, shiny, normalized UTF-8 string
78 */
79 static function cleanUp( $string ) {
80 if( NORMALIZE_ICU ) {
81 $string = self::replaceForNativeNormalize( $string );
82
83 # UnicodeString constructor fails if the string ends with a
84 # head byte. Add a junk char at the end, we'll strip it off.
85 return rtrim( utf8_normalize( $string . "\x01", self::UNORM_NFC ), "\x01" );
86 } elseif( NORMALIZE_INTL ) {
87 $string = self::replaceForNativeNormalize( $string );
88 $norm = normalizer_normalize( $string, Normalizer::FORM_C );
89 if( $norm === null || $norm === false ) {
90 # normalizer_normalize will either return false or null
91 # (depending on which doc you read) if invalid utf8 string.
92 # quickIsNFCVerify cleans up invalid sequences.
93
94 if( UtfNormal::quickIsNFCVerify( $string ) ) {
95 # if that's true, the string is actually already normal.
96 return $string;
97 } else {
98 # Now we are valid but non-normal
99 return normalizer_normalize( $string, Normalizer::FORM_C );
100 }
101 } else {
102 return $norm;
103 }
104 } elseif( UtfNormal::quickIsNFCVerify( $string ) ) {
105 # Side effect -- $string has had UTF-8 errors cleaned up.
106 return $string;
107 } else {
108 return UtfNormal::NFC( $string );
109 }
110 }
111
112 /**
113 * Convert a UTF-8 string to normal form C, canonical composition.
114 * Fast return for pure ASCII strings; some lesser optimizations for
115 * strings containing only known-good characters.
116 *
117 * @param string $string a valid UTF-8 string. Input is not validated.
118 * @return string a UTF-8 string in normal form C
119 */
120 static function toNFC( $string ) {
121 if( NORMALIZE_INTL )
122 return normalizer_normalize( $string, Normalizer::FORM_C );
123 elseif( NORMALIZE_ICU )
124 return utf8_normalize( $string, self::UNORM_NFC );
125 elseif( UtfNormal::quickIsNFC( $string ) )
126 return $string;
127 else
128 return UtfNormal::NFC( $string );
129 }
130
131 /**
132 * Convert a UTF-8 string to normal form D, canonical decomposition.
133 * Fast return for pure ASCII strings.
134 *
135 * @param string $string a valid UTF-8 string. Input is not validated.
136 * @return string a UTF-8 string in normal form D
137 */
138 static function toNFD( $string ) {
139 if( NORMALIZE_INTL )
140 return normalizer_normalize( $string, Normalizer::FORM_D );
141 elseif( NORMALIZE_ICU )
142 return utf8_normalize( $string, self::UNORM_NFD );
143 elseif( preg_match( '/[\x80-\xff]/', $string ) )
144 return UtfNormal::NFD( $string );
145 else
146 return $string;
147 }
148
149 /**
150 * Convert a UTF-8 string to normal form KC, compatibility composition.
151 * This may cause irreversible information loss, use judiciously.
152 * Fast return for pure ASCII strings.
153 *
154 * @param string $string a valid UTF-8 string. Input is not validated.
155 * @return string a UTF-8 string in normal form KC
156 */
157 static function toNFKC( $string ) {
158 if( NORMALIZE_INTL )
159 return normalizer_normalize( $string, Normalizer::FORM_KC );
160 elseif( NORMALIZE_ICU )
161 return utf8_normalize( $string, self::UNORM_NFKC );
162 elseif( preg_match( '/[\x80-\xff]/', $string ) )
163 return UtfNormal::NFKC( $string );
164 else
165 return $string;
166 }
167
168 /**
169 * Convert a UTF-8 string to normal form KD, compatibility decomposition.
170 * This may cause irreversible information loss, use judiciously.
171 * Fast return for pure ASCII strings.
172 *
173 * @param string $string a valid UTF-8 string. Input is not validated.
174 * @return string a UTF-8 string in normal form KD
175 */
176 static function toNFKD( $string ) {
177 if( NORMALIZE_INTL )
178 return normalizer_normalize( $string, Normalizer::FORM_KD );
179 elseif( NORMALIZE_ICU )
180 return utf8_normalize( $string, self::UNORM_NFKD );
181 elseif( preg_match( '/[\x80-\xff]/', $string ) )
182 return UtfNormal::NFKD( $string );
183 else
184 return $string;
185 }
186
187 /**
188 * Load the basic composition data if necessary
189 * @private
190 */
191 static function loadData() {
192 if( !isset( self::$utfCombiningClass ) ) {
193 require_once __DIR__ . '/UtfNormalData.inc';
194 }
195 }
196
197 /**
198 * Returns true if the string is _definitely_ in NFC.
199 * Returns false if not or uncertain.
200 * @param string $string a valid UTF-8 string. Input is not validated.
201 * @return bool
202 */
203 static function quickIsNFC( $string ) {
204 # ASCII is always valid NFC!
205 # If it's pure ASCII, let it through.
206 if( !preg_match( '/[\x80-\xff]/', $string ) ) return true;
207
208 UtfNormal::loadData();
209 $len = strlen( $string );
210 for( $i = 0; $i < $len; $i++ ) {
211 $c = $string[$i];
212 $n = ord( $c );
213 if( $n < 0x80 ) {
214 continue;
215 } elseif( $n >= 0xf0 ) {
216 $c = substr( $string, $i, 4 );
217 $i += 3;
218 } elseif( $n >= 0xe0 ) {
219 $c = substr( $string, $i, 3 );
220 $i += 2;
221 } elseif( $n >= 0xc0 ) {
222 $c = substr( $string, $i, 2 );
223 $i++;
224 }
225 if( isset( self::$utfCheckNFC[$c] ) ) {
226 # If it's NO or MAYBE, bail and do the slow check.
227 return false;
228 }
229 if( isset( self::$utfCombiningClass[$c] ) ) {
230 # Combining character? We might have to do sorting, at least.
231 return false;
232 }
233 }
234 return true;
235 }
236
237 /**
238 * Returns true if the string is _definitely_ in NFC.
239 * Returns false if not or uncertain.
240 * @param string $string a UTF-8 string, altered on output to be valid UTF-8 safe for XML.
241 * @return bool
242 */
243 static function quickIsNFCVerify( &$string ) {
244 # Screen out some characters that eg won't be allowed in XML
245 $string = preg_replace( '/[\x00-\x08\x0b\x0c\x0e-\x1f]/', UTF8_REPLACEMENT, $string );
246
247 # ASCII is always valid NFC!
248 # If we're only ever given plain ASCII, we can avoid the overhead
249 # of initializing the decomposition tables by skipping out early.
250 if( !preg_match( '/[\x80-\xff]/', $string ) ) return true;
251
252 static $checkit = null, $tailBytes = null, $utfCheckOrCombining = null;
253 if( !isset( $checkit ) ) {
254 # Load/build some scary lookup tables...
255 UtfNormal::loadData();
256
257 $utfCheckOrCombining = array_merge( self::$utfCheckNFC, self::$utfCombiningClass );
258
259 # Head bytes for sequences which we should do further validity checks
260 $checkit = array_flip( array_map( 'chr',
261 array( 0xc0, 0xc1, 0xe0, 0xed, 0xef,
262 0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7,
263 0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff ) ) );
264
265 # Each UTF-8 head byte is followed by a certain
266 # number of tail bytes.
267 $tailBytes = array();
268 for( $n = 0; $n < 256; $n++ ) {
269 if( $n < 0xc0 ) {
270 $remaining = 0;
271 } elseif( $n < 0xe0 ) {
272 $remaining = 1;
273 } elseif( $n < 0xf0 ) {
274 $remaining = 2;
275 } elseif( $n < 0xf8 ) {
276 $remaining = 3;
277 } elseif( $n < 0xfc ) {
278 $remaining = 4;
279 } elseif( $n < 0xfe ) {
280 $remaining = 5;
281 } else {
282 $remaining = 0;
283 }
284 $tailBytes[chr($n)] = $remaining;
285 }
286 }
287
288 # Chop the text into pure-ASCII and non-ASCII areas;
289 # large ASCII parts can be handled much more quickly.
290 # Don't chop up Unicode areas for punctuation, though,
291 # that wastes energy.
292 $matches = array();
293 preg_match_all(
294 '/([\x00-\x7f]+|[\x80-\xff][\x00-\x40\x5b-\x5f\x7b-\xff]*)/',
295 $string, $matches );
296
297 $looksNormal = true;
298 $base = 0;
299 $replace = array();
300 foreach( $matches[1] as $str ) {
301 $chunk = strlen( $str );
302
303 if( $str[0] < "\x80" ) {
304 # ASCII chunk: guaranteed to be valid UTF-8
305 # and in normal form C, so skip over it.
306 $base += $chunk;
307 continue;
308 }
309
310 # We'll have to examine the chunk byte by byte to ensure
311 # that it consists of valid UTF-8 sequences, and to see
312 # if any of them might not be normalized.
313 #
314 # Since PHP is not the fastest language on earth, some of
315 # this code is a little ugly with inner loop optimizations.
316
317 $head = '';
318 $len = $chunk + 1; # Counting down is faster. I'm *so* sorry.
319
320 for( $i = -1; --$len; ) {
321 $remaining = $tailBytes[$c = $str[++$i]];
322 if( $remaining ) {
323 # UTF-8 head byte!
324 $sequence = $head = $c;
325 do {
326 # Look for the defined number of tail bytes...
327 if( --$len && ( $c = $str[++$i] ) >= "\x80" && $c < "\xc0" ) {
328 # Legal tail bytes are nice.
329 $sequence .= $c;
330 } else {
331 if( 0 == $len ) {
332 # Premature end of string!
333 # Drop a replacement character into output to
334 # represent the invalid UTF-8 sequence.
335 $replace[] = array( UTF8_REPLACEMENT,
336 $base + $i + 1 - strlen( $sequence ),
337 strlen( $sequence ) );
338 break 2;
339 } else {
340 # Illegal tail byte; abandon the sequence.
341 $replace[] = array( UTF8_REPLACEMENT,
342 $base + $i - strlen( $sequence ),
343 strlen( $sequence ) );
344 # Back up and reprocess this byte; it may itself
345 # be a legal ASCII or UTF-8 sequence head.
346 --$i;
347 ++$len;
348 continue 2;
349 }
350 }
351 } while( --$remaining );
352
353 if( isset( $checkit[$head] ) ) {
354 # Do some more detailed validity checks, for
355 # invalid characters and illegal sequences.
356 if( $head == "\xed" ) {
357 # 0xed is relatively frequent in Korean, which
358 # abuts the surrogate area, so we're doing
359 # this check separately to speed things up.
360
361 if( $sequence >= UTF8_SURROGATE_FIRST ) {
362 # Surrogates are legal only in UTF-16 code.
363 # They are totally forbidden here in UTF-8
364 # utopia.
365 $replace[] = array( UTF8_REPLACEMENT,
366 $base + $i + 1 - strlen( $sequence ),
367 strlen( $sequence ) );
368 $head = '';
369 continue;
370 }
371 } else {
372 # Slower, but rarer checks...
373 $n = ord( $head );
374 if(
375 # "Overlong sequences" are those that are syntactically
376 # correct but use more UTF-8 bytes than are necessary to
377 # encode a character. Naïve string comparisons can be
378 # tricked into failing to see a match for an ASCII
379 # character, for instance, which can be a security hole
380 # if blacklist checks are being used.
381 ($n < 0xc2 && $sequence <= UTF8_OVERLONG_A)
382 || ($n == 0xe0 && $sequence <= UTF8_OVERLONG_B)
383 || ($n == 0xf0 && $sequence <= UTF8_OVERLONG_C)
384
385 # U+FFFE and U+FFFF are explicitly forbidden in Unicode.
386 || ($n == 0xef &&
387 ($sequence == UTF8_FFFE)
388 || ($sequence == UTF8_FFFF) )
389
390 # Unicode has been limited to 21 bits; longer
391 # sequences are not allowed.
392 || ($n >= 0xf0 && $sequence > UTF8_MAX) ) {
393
394 $replace[] = array( UTF8_REPLACEMENT,
395 $base + $i + 1 - strlen( $sequence ),
396 strlen( $sequence ) );
397 $head = '';
398 continue;
399 }
400 }
401 }
402
403 if( isset( $utfCheckOrCombining[$sequence] ) ) {
404 # If it's NO or MAYBE, we'll have to rip
405 # the string apart and put it back together.
406 # That's going to be mighty slow.
407 $looksNormal = false;
408 }
409
410 # The sequence is legal!
411 $head = '';
412 } elseif( $c < "\x80" ) {
413 # ASCII byte.
414 $head = '';
415 } elseif( $c < "\xc0" ) {
416 # Illegal tail bytes
417 if( $head == '' ) {
418 # Out of the blue!
419 $replace[] = array( UTF8_REPLACEMENT, $base + $i, 1 );
420 } else {
421 # Don't add if we're continuing a broken sequence;
422 # we already put a replacement character when we looked
423 # at the broken sequence.
424 $replace[] = array( '', $base + $i, 1 );
425 }
426 } else {
427 # Miscellaneous freaks.
428 $replace[] = array( UTF8_REPLACEMENT, $base + $i, 1 );
429 $head = '';
430 }
431 }
432 $base += $chunk;
433 }
434 if( count( $replace ) ) {
435 # There were illegal UTF-8 sequences we need to fix up.
436 $out = '';
437 $last = 0;
438 foreach( $replace as $rep ) {
439 list( $replacement, $start, $length ) = $rep;
440 if( $last < $start ) {
441 $out .= substr( $string, $last, $start - $last );
442 }
443 $out .= $replacement;
444 $last = $start + $length;
445 }
446 if( $last < strlen( $string ) ) {
447 $out .= substr( $string, $last );
448 }
449 $string = $out;
450 }
451 return $looksNormal;
452 }
453
454 # These take a string and run the normalization on them, without
455 # checking for validity or any optimization etc. Input must be
456 # VALID UTF-8!
457 /**
458 * @param $string string
459 * @return string
460 * @private
461 */
462 static function NFC( $string ) {
463 return UtfNormal::fastCompose( UtfNormal::NFD( $string ) );
464 }
465
466 /**
467 * @param $string string
468 * @return string
469 * @private
470 */
471 static function NFD( $string ) {
472 UtfNormal::loadData();
473
474 return UtfNormal::fastCombiningSort(
475 UtfNormal::fastDecompose( $string, self::$utfCanonicalDecomp ) );
476 }
477
478 /**
479 * @param $string string
480 * @return string
481 * @private
482 */
483 static function NFKC( $string ) {
484 return UtfNormal::fastCompose( UtfNormal::NFKD( $string ) );
485 }
486
487 /**
488 * @param $string string
489 * @return string
490 * @private
491 */
492 static function NFKD( $string ) {
493 if( !isset( self::$utfCompatibilityDecomp ) ) {
494 require_once 'UtfNormalDataK.inc';
495 }
496 return self::fastCombiningSort(
497 self::fastDecompose( $string, self::$utfCompatibilityDecomp ) );
498 }
499
500
501 /**
502 * Perform decomposition of a UTF-8 string into either D or KD form
503 * (depending on which decomposition map is passed to us).
504 * Input is assumed to be *valid* UTF-8. Invalid code will break.
505 * @private
506 * @param string $string valid UTF-8 string
507 * @param array $map hash of expanded decomposition map
508 * @return string a UTF-8 string decomposed, not yet normalized (needs sorting)
509 */
510 static function fastDecompose( $string, $map ) {
511 UtfNormal::loadData();
512 $len = strlen( $string );
513 $out = '';
514 for( $i = 0; $i < $len; $i++ ) {
515 $c = $string[$i];
516 $n = ord( $c );
517 if( $n < 0x80 ) {
518 # ASCII chars never decompose
519 # THEY ARE IMMORTAL
520 $out .= $c;
521 continue;
522 } elseif( $n >= 0xf0 ) {
523 $c = substr( $string, $i, 4 );
524 $i += 3;
525 } elseif( $n >= 0xe0 ) {
526 $c = substr( $string, $i, 3 );
527 $i += 2;
528 } elseif( $n >= 0xc0 ) {
529 $c = substr( $string, $i, 2 );
530 $i++;
531 }
532 if( isset( $map[$c] ) ) {
533 $out .= $map[$c];
534 continue;
535 } else {
536 if( $c >= UTF8_HANGUL_FIRST && $c <= UTF8_HANGUL_LAST ) {
537 # Decompose a hangul syllable into jamo;
538 # hardcoded for three-byte UTF-8 sequence.
539 # A lookup table would be slightly faster,
540 # but adds a lot of memory & disk needs.
541 #
542 $index = ( (ord( $c[0] ) & 0x0f) << 12
543 | (ord( $c[1] ) & 0x3f) << 6
544 | (ord( $c[2] ) & 0x3f) )
545 - UNICODE_HANGUL_FIRST;
546 $l = intval( $index / UNICODE_HANGUL_NCOUNT );
547 $v = intval( ($index % UNICODE_HANGUL_NCOUNT) / UNICODE_HANGUL_TCOUNT);
548 $t = $index % UNICODE_HANGUL_TCOUNT;
549 $out .= "\xe1\x84" . chr( 0x80 + $l ) . "\xe1\x85" . chr( 0xa1 + $v );
550 if( $t >= 25 ) {
551 $out .= "\xe1\x87" . chr( 0x80 + $t - 25 );
552 } elseif( $t ) {
553 $out .= "\xe1\x86" . chr( 0xa7 + $t );
554 }
555 continue;
556 }
557 }
558 $out .= $c;
559 }
560 return $out;
561 }
562
563 /**
564 * Sorts combining characters into canonical order. This is the
565 * final step in creating decomposed normal forms D and KD.
566 * @private
567 * @param string $string a valid, decomposed UTF-8 string. Input is not validated.
568 * @return string a UTF-8 string with combining characters sorted in canonical order
569 */
570 static function fastCombiningSort( $string ) {
571 UtfNormal::loadData();
572 $len = strlen( $string );
573 $out = '';
574 $combiners = array();
575 $lastClass = -1;
576 for( $i = 0; $i < $len; $i++ ) {
577 $c = $string[$i];
578 $n = ord( $c );
579 if( $n >= 0x80 ) {
580 if( $n >= 0xf0 ) {
581 $c = substr( $string, $i, 4 );
582 $i += 3;
583 } elseif( $n >= 0xe0 ) {
584 $c = substr( $string, $i, 3 );
585 $i += 2;
586 } elseif( $n >= 0xc0 ) {
587 $c = substr( $string, $i, 2 );
588 $i++;
589 }
590 if( isset( self::$utfCombiningClass[$c] ) ) {
591 $lastClass = self::$utfCombiningClass[$c];
592 if( isset( $combiners[$lastClass] ) ) {
593 $combiners[$lastClass] .= $c;
594 } else {
595 $combiners[$lastClass] = $c;
596 }
597 continue;
598 }
599 }
600 if( $lastClass ) {
601 ksort( $combiners );
602 $out .= implode( '', $combiners );
603 $combiners = array();
604 }
605 $out .= $c;
606 $lastClass = 0;
607 }
608 if( $lastClass ) {
609 ksort( $combiners );
610 $out .= implode( '', $combiners );
611 }
612 return $out;
613 }
614
615 /**
616 * Produces canonically composed sequences, i.e. normal form C or KC.
617 *
618 * @private
619 * @param string $string a valid UTF-8 string in sorted normal form D or KD. Input is not validated.
620 * @return string a UTF-8 string with canonical precomposed characters used where possible
621 */
622 static function fastCompose( $string ) {
623 UtfNormal::loadData();
624 $len = strlen( $string );
625 $out = '';
626 $lastClass = -1;
627 $lastHangul = 0;
628 $startChar = '';
629 $combining = '';
630 $x1 = ord(substr(UTF8_HANGUL_VBASE, 0, 1));
631 $x2 = ord(substr(UTF8_HANGUL_TEND, 0, 1));
632 for( $i = 0; $i < $len; $i++ ) {
633 $c = $string[$i];
634 $n = ord( $c );
635 if( $n < 0x80 ) {
636 # No combining characters here...
637 $out .= $startChar;
638 $out .= $combining;
639 $startChar = $c;
640 $combining = '';
641 $lastClass = 0;
642 continue;
643 } elseif( $n >= 0xf0 ) {
644 $c = substr( $string, $i, 4 );
645 $i += 3;
646 } elseif( $n >= 0xe0 ) {
647 $c = substr( $string, $i, 3 );
648 $i += 2;
649 } elseif( $n >= 0xc0 ) {
650 $c = substr( $string, $i, 2 );
651 $i++;
652 }
653 $pair = $startChar . $c;
654 if( $n > 0x80 ) {
655 if( isset( self::$utfCombiningClass[$c] ) ) {
656 # A combining char; see what we can do with it
657 $class = self::$utfCombiningClass[$c];
658 if( !empty( $startChar ) &&
659 $lastClass < $class &&
660 $class > 0 &&
661 isset( self::$utfCanonicalComp[$pair] ) ) {
662 $startChar = self::$utfCanonicalComp[$pair];
663 $class = 0;
664 } else {
665 $combining .= $c;
666 }
667 $lastClass = $class;
668 $lastHangul = 0;
669 continue;
670 }
671 }
672 # New start char
673 if( $lastClass == 0 ) {
674 if( isset( self::$utfCanonicalComp[$pair] ) ) {
675 $startChar = self::$utfCanonicalComp[$pair];
676 $lastHangul = 0;
677 continue;
678 }
679 if( $n >= $x1 && $n <= $x2 ) {
680 # WARNING: Hangul code is painfully slow.
681 # I apologize for this ugly, ugly code; however
682 # performance is even more teh suck if we call
683 # out to nice clean functions. Lookup tables are
684 # marginally faster, but require a lot of space.
685 #
686 if( $c >= UTF8_HANGUL_VBASE &&
687 $c <= UTF8_HANGUL_VEND &&
688 $startChar >= UTF8_HANGUL_LBASE &&
689 $startChar <= UTF8_HANGUL_LEND ) {
690 #
691 #$lIndex = utf8ToCodepoint( $startChar ) - UNICODE_HANGUL_LBASE;
692 #$vIndex = utf8ToCodepoint( $c ) - UNICODE_HANGUL_VBASE;
693 $lIndex = ord( $startChar[2] ) - 0x80;
694 $vIndex = ord( $c[2] ) - 0xa1;
695
696 $hangulPoint = UNICODE_HANGUL_FIRST +
697 UNICODE_HANGUL_TCOUNT *
698 (UNICODE_HANGUL_VCOUNT * $lIndex + $vIndex);
699
700 # Hardcode the limited-range UTF-8 conversion:
701 $startChar = chr( $hangulPoint >> 12 & 0x0f | 0xe0 ) .
702 chr( $hangulPoint >> 6 & 0x3f | 0x80 ) .
703 chr( $hangulPoint & 0x3f | 0x80 );
704 $lastHangul = 0;
705 continue;
706 } elseif( $c >= UTF8_HANGUL_TBASE &&
707 $c <= UTF8_HANGUL_TEND &&
708 $startChar >= UTF8_HANGUL_FIRST &&
709 $startChar <= UTF8_HANGUL_LAST &&
710 !$lastHangul ) {
711 # $tIndex = utf8ToCodepoint( $c ) - UNICODE_HANGUL_TBASE;
712 $tIndex = ord( $c[2] ) - 0xa7;
713 if( $tIndex < 0 ) $tIndex = ord( $c[2] ) - 0x80 + (0x11c0 - 0x11a7);
714
715 # Increment the code point by $tIndex, without
716 # the function overhead of decoding and recoding UTF-8
717 #
718 $tail = ord( $startChar[2] ) + $tIndex;
719 if( $tail > 0xbf ) {
720 $tail -= 0x40;
721 $mid = ord( $startChar[1] ) + 1;
722 if( $mid > 0xbf ) {
723 $startChar[0] = chr( ord( $startChar[0] ) + 1 );
724 $mid -= 0x40;
725 }
726 $startChar[1] = chr( $mid );
727 }
728 $startChar[2] = chr( $tail );
729
730 # If there's another jamo char after this, *don't* try to merge it.
731 $lastHangul = 1;
732 continue;
733 }
734 }
735 }
736 $out .= $startChar;
737 $out .= $combining;
738 $startChar = $c;
739 $combining = '';
740 $lastClass = 0;
741 $lastHangul = 0;
742 }
743 $out .= $startChar . $combining;
744 return $out;
745 }
746
747 /**
748 * This is just used for the benchmark, comparing how long it takes to
749 * interate through a string without really doing anything of substance.
750 * @param $string string
751 * @return string
752 */
753 static function placebo( $string ) {
754 $len = strlen( $string );
755 $out = '';
756 for( $i = 0; $i < $len; $i++ ) {
757 $out .= $string[$i];
758 }
759 return $out;
760 }
761 /**
762 * Function to replace some characters that we don't want
763 * but most of the native normalize functions keep.
764 *
765 * @param string $string The string
766 * @return String String with the character codes replaced.
767 */
768 private static function replaceForNativeNormalize( $string ) {
769 $string = preg_replace(
770 '/[\x00-\x08\x0b\x0c\x0e-\x1f]/',
771 UTF8_REPLACEMENT,
772 $string );
773 $string = str_replace( UTF8_FFFE, UTF8_REPLACEMENT, $string );
774 $string = str_replace( UTF8_FFFF, UTF8_REPLACEMENT, $string );
775 return $string;
776 }
777 }
MediaWiki Core