2 # Copyright (C) 2004 Brion Vibber <brion@pobox.com>
3 # http://www.mediawiki.org/
5 # This program is free software; you can redistribute it and/or modify
6 # it under the terms of the GNU General Public License as published by
7 # the Free Software Foundation; either version 2 of the License, or
8 # (at your option) any later version.
10 # This program is distributed in the hope that it will be useful,
11 # but WITHOUT ANY WARRANTY; without even the implied warranty of
12 # MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
13 # GNU General Public License for more details.
15 # You should have received a copy of the GNU General Public License along
16 # with this program; if not, write to the Free Software Foundation, Inc.,
17 # 59 Temple Place - Suite 330, Boston, MA 02111-1307, USA.
18 # http://www.gnu.org/copyleft/gpl.html
21 * Unicode normalization routines for working with UTF-8 strings.
22 * Currently assumes that input strings are valid UTF-8!
24 * Not as fast as I'd like, but should be usable for most purposes.
25 * UtfNormal::toNFC() will bail early if given ASCII text or text
26 * it can quickly deterimine is already normalized.
28 * All functions can be called static.
30 * See description of forms at http://www.unicode.org/reports/tr15/
36 require_once 'UtfNormalUtil.php';
37 require_once 'UtfNormalData.inc';
39 # Load compatibility decompositions on demand if they are needed.
40 global $utfCompatibilityDecomp;
41 $utfCompatibilityDecomp = NULL;
43 define( 'UNICODE_HANGUL_FIRST', 0xac00 );
44 define( 'UNICODE_HANGUL_LAST', 0xd7a3 );
46 define( 'UNICODE_HANGUL_LBASE', 0x1100 );
47 define( 'UNICODE_HANGUL_VBASE', 0x1161 );
48 define( 'UNICODE_HANGUL_TBASE', 0x11a7 );
50 define( 'UNICODE_HANGUL_LCOUNT', 19 );
51 define( 'UNICODE_HANGUL_VCOUNT', 21 );
52 define( 'UNICODE_HANGUL_TCOUNT', 28 );
53 define( 'UNICODE_HANGUL_NCOUNT', UNICODE_HANGUL_VCOUNT
* UNICODE_HANGUL_TCOUNT
);
55 define( 'UNICODE_HANGUL_LEND', UNICODE_HANGUL_LBASE + UNICODE_HANGUL_LCOUNT
- 1 );
56 define( 'UNICODE_HANGUL_VEND', UNICODE_HANGUL_VBASE + UNICODE_HANGUL_VCOUNT
- 1 );
57 define( 'UNICODE_HANGUL_TEND', UNICODE_HANGUL_TBASE + UNICODE_HANGUL_TCOUNT
- 1 );
59 define( 'UNICODE_SURROGATE_FIRST', 0xd800 );
60 define( 'UNICODE_SURROGATE_LAST', 0xdfff );
61 define( 'UNICODE_MAX', 0x10ffff );
62 define( 'UNICODE_REPLACEMENT', 0xfffd );
65 define( 'UTF8_HANGUL_FIRST', codepointToUtf8( UNICODE_HANGUL_FIRST
) );
66 define( 'UTF8_HANGUL_LAST', codepointToUtf8( UNICODE_HANGUL_LAST
) );
68 define( 'UTF8_HANGUL_LBASE', codepointToUtf8( UNICODE_HANGUL_LBASE
) );
69 define( 'UTF8_HANGUL_VBASE', codepointToUtf8( UNICODE_HANGUL_VBASE
) );
70 define( 'UTF8_HANGUL_TBASE', codepointToUtf8( UNICODE_HANGUL_TBASE
) );
72 define( 'UTF8_HANGUL_LEND', codepointToUtf8( UNICODE_HANGUL_LEND
) );
73 define( 'UTF8_HANGUL_VEND', codepointToUtf8( UNICODE_HANGUL_VEND
) );
74 define( 'UTF8_HANGUL_TEND', codepointToUtf8( UNICODE_HANGUL_TEND
) );
76 define( 'UTF8_SURROGATE_FIRST', codepointToUtf8( UNICODE_SURROGATE_FIRST
) );
77 define( 'UTF8_SURROGATE_LAST', codepointToUtf8( UNICODE_SURROGATE_LAST
) );
78 define( 'UTF8_MAX', codepointToUtf8( UNICODE_MAX
) );
79 define( 'UTF8_REPLACEMENT', codepointToUtf8( UNICODE_REPLACEMENT
) );
80 #define( 'UTF8_REPLACEMENT', '!' );
82 define( 'UTF8_OVERLONG_A', "\xc1\xbf" );
83 define( 'UTF8_OVERLONG_B', "\xe0\x9f\xbf" );
84 define( 'UTF8_OVERLONG_C', "\xf0\x8f\xbf\xbf" );
86 # These two ranges are illegal
87 define( 'UTF8_FDD0', codepointToUtf8( 0xfdd0 ) );
88 define( 'UTF8_FDEF', codepointToUtf8( 0xfdef ) );
89 define( 'UTF8_FFFE', codepointToUtf8( 0xfffe ) );
90 define( 'UTF8_FFFF', codepointToUtf8( 0xffff ) );
92 define( 'UTF8_HEAD', false );
93 define( 'UTF8_TAIL', true );
97 * For using the ICU wrapper
99 define( 'UNORM_NONE', 1 );
100 define( 'UNORM_NFD', 2 );
101 define( 'UNORM_NFKD', 3 );
102 define( 'UNORM_NFC', 4 );
103 define( 'UNORM_DEFAULT', UNORM_NFC
);
104 define( 'UNORM_NFKC', 5 );
105 define( 'UNORM_FCD', 6 );
107 define( 'NORMALIZE_ICU', function_exists( 'utf8_normalize' ) );
115 * The ultimate convenience function! Clean up invalid UTF-8 sequences,
116 * and convert to normal form C, canonical composition.
118 * Fast return for pure ASCII strings; some lesser optimizations for
119 * strings containing only known-good characters. Not as fast as toNFC().
121 * @param string $string a UTF-8 string
122 * @return string a clean, shiny, normalized UTF-8 string
124 function cleanUp( $string ) {
125 if( UtfNormal
::quickIsNFCVerify( $string ) )
128 return UtfNormal
::NFC( $string );
132 * Convert a UTF-8 string to normal form C, canonical composition.
133 * Fast return for pure ASCII strings; some lesser optimizations for
134 * strings containing only known-good characters.
136 * @param string $string a valid UTF-8 string. Input is not validated.
137 * @return string a UTF-8 string in normal form C
139 function toNFC( $string ) {
141 return utf8_normalize( $string, UNORM_NFC
);
142 elseif( UtfNormal
::quickIsNFC( $string ) )
145 return UtfNormal
::NFC( $string );
149 * Convert a UTF-8 string to normal form D, canonical decomposition.
150 * Fast return for pure ASCII strings.
152 * @param string $string a valid UTF-8 string. Input is not validated.
153 * @return string a UTF-8 string in normal form D
155 function toNFD( $string ) {
157 return utf8_normalize( $string, UNORM_NFD
);
158 elseif( preg_match( '/[\x80-\xff]/', $string ) )
159 return UtfNormal
::NFD( $string );
165 * Convert a UTF-8 string to normal form KC, compatibility composition.
166 * This may cause irreversible information loss, use judiciously.
167 * Fast return for pure ASCII strings.
169 * @param string $string a valid UTF-8 string. Input is not validated.
170 * @return string a UTF-8 string in normal form KC
172 function toNFKC( $string ) {
174 return utf8_normalize( $string, UNORM_NFKC
);
175 elseif( preg_match( '/[\x80-\xff]/', $string ) )
176 return UtfNormal
::NFKC( $string );
182 * Convert a UTF-8 string to normal form KD, compatibility decomposition.
183 * This may cause irreversible information loss, use judiciously.
184 * Fast return for pure ASCII strings.
186 * @param string $string a valid UTF-8 string. Input is not validated.
187 * @return string a UTF-8 string in normal form KD
189 function toNFKD( $string ) {
191 return utf8_normalize( $string, UNORM_NFKD
);
192 elseif( preg_match( '/[\x80-\xff]/', $string ) )
193 return UtfNormal
::NFKD( $string );
199 * Returns true if the string is _definitely_ in NFC.
200 * Returns false if not or uncertain.
201 * @param string $string a valid UTF-8 string. Input is not validated.
204 function quickIsNFC( $string ) {
205 # ASCII is always valid NFC!
206 # If it's pure ASCII, let it through.
207 if( !preg_match( '/[\x80-\xff]/', $string ) ) return true;
209 global $utfCheckNFC, $utfCombiningClass;
210 $len = strlen( $string );
211 for( $i = 0; $i < $len; $i++
) {
216 } elseif( $n >= 0xf0 ) {
217 $c = substr( $string, $i, 4 );
219 } elseif( $n >= 0xe0 ) {
220 $c = substr( $string, $i, 3 );
222 } elseif( $n >= 0xc0 ) {
223 $c = substr( $string, $i, 2 );
226 if( isset( $utfCheckNFC[$c] ) ) {
227 # If it's NO or MAYBE, bail and do the slow check.
230 if( isset( $utfCombiningClass[$c] ) ) {
231 # Combining character? We might have to do sorting, at least.
239 * Returns true if the string is _definitely_ in NFC.
240 * Returns false if not or uncertain.
241 * @param string $string a UTF-8 string, altered on output to be valid UTF-8 safe for XML.
244 function quickIsNFCVerify( &$string ) {
245 # ASCII is always valid NFC!
246 if( !preg_match( '/[\x80-\xff]/', $string ) ) return true;
248 global $utfCheckNFC, $utfCombiningClass;
249 $len = strlen( $string );
256 for( $i = 0; $i < $len; $i++
) {
259 if( $state == UTF8_TAIL
) {
260 if( $n >= 0x80 && $n < 0xc0 ) {
262 if( --$remaining == 0 ) {
263 if( ($sequence >= UTF8_SURROGATE_FIRST
264 && $sequence <= UTF8_SURROGATE_LAST
)
265 ||
($head == 0xc0 && $sequence <= UTF8_OVERLONG_A
)
266 ||
($head == 0xc1 && $sequence <= UTF8_OVERLONG_A
)
267 ||
($head == 0xe0 && $sequence <= UTF8_OVERLONG_B
)
268 ||
($head == 0xf0 && $sequence <= UTF8_OVERLONG_C
)
269 ||
($sequence >= UTF8_FDD0
&& $sequence <= UTF8_FDEF
)
270 ||
($sequence == UTF8_FFFE
)
271 ||
($sequence == UTF8_FFFF
)
272 ||
($sequence > UTF8_MAX
) ) {
273 $out .= UTF8_REPLACEMENT
;
277 if( isset( $utfCheckNFC[$sequence] ) ||
278 isset( $utfCombiningClass[$sequence] ) ) {
279 # If it's NO or MAYBE, we'll have to do the slow check.
280 $looksNormal = false;
288 # Not a valid tail byte! DIscard the char we've been building.
289 #printf ("Invalid '%x' in tail with %d remaining bytes\n", $n, $remaining );
291 $out .= UTF8_REPLACEMENT
;
294 $out .= UTF8_REPLACEMENT
;
295 } elseif( $n == 0x0a ) {
297 } elseif( $n < 0x0d ) {
298 $out .= UTF8_REPLACEMENT
;
299 } elseif( $n == 0x0d ) {
301 } elseif( $n < 0x20 ) {
302 $out .= UTF8_REPLACEMENT
;
303 } elseif( $n < 0x80 ) {
305 } elseif( $n < 0xc0 ) {
306 # illegal tail bytes or head byte of overlong sequence
307 if( $head == 0 ) $out .= UTF8_REPLACEMENT
;
308 } elseif( $n < 0xe0 ) {
313 } elseif( $n < 0xf0 ) {
318 } elseif( $n < 0xf8 ) {
323 } elseif( $n < 0xfc ) {
328 } elseif( $n < 0xfe ) {
334 $out .= UTF8_REPLACEMENT
;
337 if( $state == UTF8_TAIL
) {
338 $out .= UTF8_REPLACEMENT
;
344 # These take a string and run the normalization on them, without
345 # checking for validity or any optimization etc. Input must be
348 * @param string $string
352 function NFC( $string ) {
353 return $out = UtfNormal
::fastCompose( UtfNormal
::NFD( $string ) );
357 * @param string $string
361 function NFD( $string ) {
362 global $utfCanonicalDecomp;
363 return UtfNormal
::fastCombiningSort(
364 UtfNormal
::fastDecompose( $string, $utfCanonicalDecomp ) );
368 * @param string $string
372 function NFKC( $string ) {
373 return UtfNormal
::fastCompose( UtfNormal
::NFKD( $string ) );
377 * @param string $string
381 function NFKD( $string ) {
382 global $utfCompatibilityDecomp;
383 if( !isset( $utfCompatibilityDecomp ) ) {
384 require_once( 'UtfNormalDataK.inc' );
386 return UtfNormal
::fastCombiningSort(
387 UtfNormal
::fastDecompose( $string, $utfCompatibilityDecomp ) );
392 * Perform decomposition of a UTF-8 string into either D or KD form
393 * (depending on which decomposition map is passed to us).
394 * Input is assumed to be *valid* UTF-8. Invalid code will break.
396 * @param string &$string Valid UTF-8 string
397 * @param array &$map hash of expanded decomposition map
398 * @return string a UTF-8 string decomposed, not yet normalized (needs sorting)
400 function fastDecompose( &$string, &$map ) {
401 $len = strlen( $string );
403 for( $i = 0; $i < $len; $i++
) {
407 # ASCII chars never decompose
411 } elseif( $n >= 0xf0 ) {
412 $c = substr( $string, $i, 4 );
414 } elseif( $n >= 0xe0 ) {
415 $c = substr( $string, $i, 3 );
417 } elseif( $n >= 0xc0 ) {
418 $c = substr( $string, $i, 2 );
421 if( isset( $map[$c] ) ) {
424 if( $c >= UTF8_HANGUL_FIRST
&& $c <= UTF8_HANGUL_LAST
) {
425 $out .= UtfNormal
::decomposeHangul( $c );
435 * Decompose a Hangul syllable character into its constituent jamo.
437 * @param int $c Unicode code point of the character
438 * @return string a UTF-8 string containing a sequence of jamo
440 function decomposeHangul( $c ) {
441 $codepoint = utf8ToCodepoint( $c );
442 $index = $codepoint - UNICODE_HANGUL_FIRST
;
443 $l = IntVal( $index / UNICODE_HANGUL_NCOUNT
);
444 $v = IntVal( ($index % UNICODE_HANGUL_NCOUNT
) / UNICODE_HANGUL_TCOUNT
);
445 $t = $index % UNICODE_HANGUL_TCOUNT
;
446 $out = codepointToUtf8( $l + UNICODE_HANGUL_LBASE
);
447 $out .= codepointToUtf8( $v + UNICODE_HANGUL_VBASE
);
448 if( $t ) $out .= codepointToUtf8( $t + UNICODE_HANGUL_TBASE
);
453 * Sorts combining characters into canonical order. This is the
454 * final step in creating decomposed normal forms D and KD.
456 * @param string $string a valid, decomposed UTF-8 string. Input is not validated.
457 * @return string a UTF-8 string with combining characters sorted in canonical order
459 function fastCombiningSort( $string ) {
460 global $utfCombiningClass;
462 while( $replacedCount > 0 ) {
464 $len = strlen( $string );
468 for( $i = 0; $i < $len; $i++
) {
472 $c = substr( $string, $i, 4 );
474 } elseif( $n >= 0xe0 ) {
475 $c = substr( $string, $i, 3 );
477 } elseif( $n >= 0xc0 ) {
478 $c = substr( $string, $i, 2 );
481 $class = isset( $utfCombiningClass[$c] ) ?
$utfCombiningClass[$c] : 0;
482 if( $lastClass == -1 ) {
486 } elseif( $lastClass > $class && $class > 0 ) {
487 # Swap -- put this one on the stack
503 * Produces canonically composed sequences, i.e. normal form C or KC.
506 * @param string $string a valid UTF-8 string in sorted normal form D or KD. Input is not validated.
507 * @return string a UTF-8 string with canonical precomposed characters used where possible
509 function fastCompose( $string ) {
510 global $utfCanonicalComp, $utfCombiningClass;
511 $len = strlen( $string );
516 for( $i = 0; $i < $len; $i++
) {
520 $c = substr( $string, $i, 4 );
522 } elseif( $n >= 0xe0 ) {
523 $c = substr( $string, $i, 3 );
525 } elseif( $n >= 0xc0 ) {
526 $c = substr( $string, $i, 2 );
529 $class = isset( $utfCombiningClass[$c] ) ?
$utfCombiningClass[$c] : 0;
530 $pair = $startChar . $c;
531 if( empty( $utfCombiningClass[$c] ) ) {
533 if( $lastClass == 0 && isset( $utfCanonicalComp[$pair] ) ) {
534 $startChar = $utfCanonicalComp[$pair];
535 } elseif( $lastClass == 0 &&
536 $c >= UTF8_HANGUL_VBASE
&&
537 $c <= UTF8_HANGUL_VEND
&&
538 $startChar >= UTF8_HANGUL_LBASE
&&
539 $startChar <= UTF8_HANGUL_LEND
) {
540 $lIndex = utf8ToCodepoint( $startChar ) - UNICODE_HANGUL_LBASE
;
541 $vIndex = utf8ToCodepoint( $c ) - UNICODE_HANGUL_VBASE
;
542 $hangulPoint = UNICODE_HANGUL_FIRST +
543 UNICODE_HANGUL_TCOUNT
*
544 (UNICODE_HANGUL_VCOUNT
* $lIndex +
$vIndex);
545 $startChar = codepointToUtf8( $hangulPoint );
546 } elseif( $lastClass == 0 &&
547 $c >= UTF8_HANGUL_TBASE
&&
548 $c <= UTF8_HANGUL_TEND
&&
549 $startChar >= UTF8_HANGUL_FIRST
&&
550 $startChar <= UTF8_HANGUL_LAST
) {
551 $tIndex = utf8ToCodepoint( $c ) - UNICODE_HANGUL_TBASE
;
552 $hangulPoint = utf8ToCodepoint( $startChar ) +
$tIndex;
553 $startChar = codepointToUtf8( $hangulPoint );
561 # A combining char; see what we can do with it
562 if( !empty( $startChar ) &&
563 $lastClass < $class &&
565 isset( $utfCanonicalComp[$pair] ) ) {
566 $startChar = $utfCanonicalComp[$pair];
574 $out .= $startChar . $combining;
579 * This is just used for the benchmark, comparing how long it takes to
580 * interate through a string without really doing anything of substance.
581 * @param string $string
584 function placebo( $string ) {
585 $len = strlen( $string );
587 for( $i = 0; $i < $len; $i++
) {