includes/normal/UtfNormal.php

   1 <?php
   2 # Copyright (C) 2004 Brion Vibber <brion@pobox.com>
   3 # http://www.mediawiki.org/
   4 #
   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.
   9 #
  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.
  14 #
  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
  19
  20 /**
  21  * Unicode normalization routines for working with UTF-8 strings.
  22  * Currently assumes that input strings are valid UTF-8!
  23  *
  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.
  27  *
  28  * All functions can be called static.
  29  *
  30  * See description of forms at http://www.unicode.org/reports/tr15/
  31  *
  32  * @package UtfNormal
  33  */
  34
  35 /** */
  36 require_once 'UtfNormalUtil.php';
  37
  38 global $utfCombiningClass, $utfCanonicalComp, $utfCanonicalDecomp;
  39 $utfCombiningClass = NULL;
  40 $utfCanonicalComp = NULL;
  41 $utfCanonicalDecomp = NULL;
  42
  43 # Load compatibility decompositions on demand if they are needed.
  44 global $utfCompatibilityDecomp;
  45 $utfCompatibilityDecomp = NULL;
  46
  47 define( 'UNICODE_HANGUL_FIRST', 0xac00 );
  48 define( 'UNICODE_HANGUL_LAST',  0xd7a3 );
  49
  50 define( 'UNICODE_HANGUL_LBASE', 0x1100 );
  51 define( 'UNICODE_HANGUL_VBASE', 0x1161 );
  52 define( 'UNICODE_HANGUL_TBASE', 0x11a7 );
  53
  54 define( 'UNICODE_HANGUL_LCOUNT', 19 );
  55 define( 'UNICODE_HANGUL_VCOUNT', 21 );
  56 define( 'UNICODE_HANGUL_TCOUNT', 28 );
  57 define( 'UNICODE_HANGUL_NCOUNT', UNICODE_HANGUL_VCOUNT * UNICODE_HANGUL_TCOUNT );
  58
  59 define( 'UNICODE_HANGUL_LEND', UNICODE_HANGUL_LBASE + UNICODE_HANGUL_LCOUNT - 1 );
  60 define( 'UNICODE_HANGUL_VEND', UNICODE_HANGUL_VBASE + UNICODE_HANGUL_VCOUNT - 1 );
  61 define( 'UNICODE_HANGUL_TEND', UNICODE_HANGUL_TBASE + UNICODE_HANGUL_TCOUNT - 1 );
  62
  63 define( 'UNICODE_SURROGATE_FIRST', 0xd800 );
  64 define( 'UNICODE_SURROGATE_LAST', 0xdfff );
  65 define( 'UNICODE_MAX', 0x10ffff );
  66 define( 'UNICODE_REPLACEMENT', 0xfffd );
  67
  68
  69 define( 'UTF8_HANGUL_FIRST', codepointToUtf8( UNICODE_HANGUL_FIRST ) );
  70 define( 'UTF8_HANGUL_LAST', codepointToUtf8( UNICODE_HANGUL_LAST ) );
  71
  72 define( 'UTF8_HANGUL_LBASE', codepointToUtf8( UNICODE_HANGUL_LBASE ) );
  73 define( 'UTF8_HANGUL_VBASE', codepointToUtf8( UNICODE_HANGUL_VBASE ) );
  74 define( 'UTF8_HANGUL_TBASE', codepointToUtf8( UNICODE_HANGUL_TBASE ) );
  75
  76 define( 'UTF8_HANGUL_LEND', codepointToUtf8( UNICODE_HANGUL_LEND ) );
  77 define( 'UTF8_HANGUL_VEND', codepointToUtf8( UNICODE_HANGUL_VEND ) );
  78 define( 'UTF8_HANGUL_TEND', codepointToUtf8( UNICODE_HANGUL_TEND ) );
  79
  80 define( 'UTF8_SURROGATE_FIRST', codepointToUtf8( UNICODE_SURROGATE_FIRST ) );
  81 define( 'UTF8_SURROGATE_LAST', codepointToUtf8( UNICODE_SURROGATE_LAST ) );
  82 define( 'UTF8_MAX', codepointToUtf8( UNICODE_MAX ) );
  83 define( 'UTF8_REPLACEMENT', codepointToUtf8( UNICODE_REPLACEMENT ) );
  84 #define( 'UTF8_REPLACEMENT', '!' );
  85
  86 define( 'UTF8_OVERLONG_A', "\xc1\xbf" );
  87 define( 'UTF8_OVERLONG_B', "\xe0\x9f\xbf" );
  88 define( 'UTF8_OVERLONG_C', "\xf0\x8f\xbf\xbf" );
  89
  90 # These two ranges are illegal
  91 define( 'UTF8_FDD0', codepointToUtf8( 0xfdd0 ) );
  92 define( 'UTF8_FDEF', codepointToUtf8( 0xfdef ) );
  93 define( 'UTF8_FFFE', codepointToUtf8( 0xfffe ) );
  94 define( 'UTF8_FFFF', codepointToUtf8( 0xffff ) );
  95
  96 define( 'UTF8_HEAD', false );
  97 define( 'UTF8_TAIL', true );
  98
  99
 100 /**
 101  * For using the ICU wrapper
 102  */
 103 define( 'UNORM_NONE', 1 );
 104 define( 'UNORM_NFD',  2 );
 105 define( 'UNORM_NFKD', 3 );
 106 define( 'UNORM_NFC',  4 );
 107 define( 'UNORM_DEFAULT', UNORM_NFC );
 108 define( 'UNORM_NFKC', 5 );
 109 define( 'UNORM_FCD',  6 );
 110
 111 define( 'NORMALIZE_ICU', function_exists( 'utf8_normalize' ) );
 112
 113 /**
 114  *
 115  * @package MediaWiki
 116  */
 117 class UtfNormal {
 118         /**
 119          * The ultimate convenience function! Clean up invalid UTF-8 sequences,
 120          * and convert to normal form C, canonical composition.
 121          *
 122          * Fast return for pure ASCII strings; some lesser optimizations for
 123          * strings containing only known-good characters. Not as fast as toNFC().
 124          *
 125          * @param string $string a UTF-8 string
 126          * @return string a clean, shiny, normalized UTF-8 string
 127          */
 128         function cleanUp( $string ) {
 129                 if( UtfNormal::quickIsNFCVerify( $string ) )
 130                         return $string;
 131                 else
 132                         return UtfNormal::NFC( $string );
 133         }
 134
 135         /**
 136          * Convert a UTF-8 string to normal form C, canonical composition.
 137          * Fast return for pure ASCII strings; some lesser optimizations for
 138          * strings containing only known-good characters.
 139          *
 140          * @param string $string a valid UTF-8 string. Input is not validated.
 141          * @return string a UTF-8 string in normal form C
 142          */
 143         function toNFC( $string ) {
 144                 if( NORMALIZE_ICU )
 145                         return utf8_normalize( $string, UNORM_NFC );
 146                 elseif( UtfNormal::quickIsNFC( $string ) )
 147                         return $string;
 148                 else
 149                         return UtfNormal::NFC( $string );
 150         }
 151
 152         /**
 153          * Convert a UTF-8 string to normal form D, canonical decomposition.
 154          * Fast return for pure ASCII strings.
 155          *
 156          * @param string $string a valid UTF-8 string. Input is not validated.
 157          * @return string a UTF-8 string in normal form D
 158          */
 159         function toNFD( $string ) {
 160                 if( NORMALIZE_ICU )
 161                         return utf8_normalize( $string, UNORM_NFD );
 162                 elseif( preg_match( '/[\x80-\xff]/', $string ) )
 163                         return UtfNormal::NFD( $string );
 164                 else
 165                         return $string;
 166         }
 167
 168         /**
 169          * Convert a UTF-8 string to normal form KC, compatibility composition.
 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 KC
 175          */
 176         function toNFKC( $string ) {
 177                 if( NORMALIZE_ICU )
 178                         return utf8_normalize( $string, UNORM_NFKC );
 179                 elseif( preg_match( '/[\x80-\xff]/', $string ) )
 180                         return UtfNormal::NFKC( $string );
 181                 else
 182                         return $string;
 183         }
 184
 185         /**
 186          * Convert a UTF-8 string to normal form KD, compatibility decomposition.
 187          * This may cause irreversible information loss, use judiciously.
 188          * Fast return for pure ASCII strings.
 189          *
 190          * @param string $string a valid UTF-8 string. Input is not validated.
 191          * @return string a UTF-8 string in normal form KD
 192          */
 193         function toNFKD( $string ) {
 194                 if( NORMALIZE_ICU )
 195                         return utf8_normalize( $string, UNORM_NFKD );
 196                 elseif( preg_match( '/[\x80-\xff]/', $string ) )
 197                         return UtfNormal::NFKD( $string );
 198                 else
 199                         return $string;
 200         }
 201
 202         /**
 203          * Load the basic composition data if necessary
 204          * @access private
 205          */
 206         function loadData() {
 207                 global $utfCombiningClass, $utfCanonicalComp, $utfCanonicalDecomp;
 208                 if( !isset( $utfCombiningClass ) ) {
 209                         require_once( 'UtfNormalData.inc' );
 210                 }
 211         }
 212
 213         /**
 214          * Returns true if the string is _definitely_ in NFC.
 215          * Returns false if not or uncertain.
 216          * @param string $string a valid UTF-8 string. Input is not validated.
 217          * @return bool
 218          */
 219         function quickIsNFC( $string ) {
 220                 # ASCII is always valid NFC!
 221                 # If it's pure ASCII, let it through.
 222                 if( !preg_match( '/[\x80-\xff]/', $string ) ) return true;
 223
 224                 UtfNormal::loadData();
 225                 global $utfCheckNFC, $utfCombiningClass;
 226                 $len = strlen( $string );
 227                 for( $i = 0; $i < $len; $i++ ) {
 228                         $c = $string{$i};
 229                         $n = ord( $c );
 230                         if( $n < 0x80 ) {
 231                                 continue;
 232                         } elseif( $n >= 0xf0 ) {
 233                                 $c = substr( $string, $i, 4 );
 234                                 $i += 3;
 235                         } elseif( $n >= 0xe0 ) {
 236                                 $c = substr( $string, $i, 3 );
 237                                 $i += 2;
 238                         } elseif( $n >= 0xc0 ) {
 239                                 $c = substr( $string, $i, 2 );
 240                                 $i++;
 241                         }
 242                         if( isset( $utfCheckNFC[$c] ) ) {
 243                                 # If it's NO or MAYBE, bail and do the slow check.
 244                                 return false;
 245                         }
 246                         if( isset( $utfCombiningClass[$c] ) ) {
 247                                 # Combining character? We might have to do sorting, at least.
 248                                 return false;
 249                         }
 250                 }
 251                 return true;
 252         }
 253
 254         /**
 255          * Returns true if the string is _definitely_ in NFC.
 256          * Returns false if not or uncertain.
 257          * @param string $string a UTF-8 string, altered on output to be valid UTF-8 safe for XML.
 258          */
 259         function quickIsNFCVerify( &$string ) {
 260                 # Screen out some characters that eg won't be allowed in XML
 261                 $string = preg_replace( '/[\x00-\x08\x0b\x0c\x0e-\x1f]/', UTF8_REPLACEMENT, $string );
 262
 263                 # ASCII is always valid NFC!
 264                 # If we're only ever given plain ASCII, we can avoid the overhead
 265                 # of initializing the decomposition tables by skipping out early.
 266                 if( !preg_match( '/[\x80-\xff]/', $string ) ) return true;
 267
 268                 UtfNormal::loadData();
 269                 global $utfCheckNFC, $utfCombiningClass;
 270
 271                 static $checkit = null, $tailBytes = null, $utfCheckOrCombining = null;
 272                 if( !isset( $checkit ) ) {
 273                         $utfCheckOrCombining = array_merge( $utfCheckNFC, $utfCombiningClass );
 274
 275                         # Head bytes for sequences which we should do further validity checks
 276                         $checkit = array_flip( array_map( 'chr',
 277                                         array( 0xc0, 0xc1, 0xe0, 0xed, 0xef,
 278                                                    0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7,
 279                                                    0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff ) ) );
 280
 281                         $tailBytes = array();
 282                         for( $n = 0; $n < 256; $n++ ) {
 283                                 if( $n < 0xc0 ) {
 284                                         $remaining = 0;
 285                                 } elseif( $n < 0xe0 ) {
 286                                         $remaining = 1;
 287                                 } elseif( $n < 0xf0 ) {
 288                                         $remaining = 2;
 289                                 } elseif( $n < 0xf8 ) {
 290                                         $remaining = 3;
 291                                 } elseif( $n < 0xfc ) {
 292                                         $remaining = 4;
 293                                 } elseif( $n < 0xfe ) {
 294                                         $remaining = 5;
 295                                 } else {
 296                                         $remaining = 0;
 297                                 }
 298                                 $tailBytes[chr($n)] = $remaining;
 299                         }
 300                 }
 301
 302                 # Chop the text into pure-ASCII and non-ASCII areas;
 303                 # large ASCII parts can be handled much more quickly.
 304                 # Don't chop up Unicode areas for punctuation, though,
 305                 # that wastes energy.
 306                 preg_match_all(
 307                         '/([\x00-\x7f]+|[\x80-\xff][\x00-\x40\x5b-\x5f\x7b-\xff]*)/',
 308                         $string, $matches );
 309
 310                 ob_start();
 311                 $looksNormal = true;
 312                 foreach( $matches[1] as $str ) {
 313                         if( $str{0} < "\x80" ) {
 314                                 # ASCII chunk: guaranteed to be valid UTF-8
 315                                 # and in normal form C, so output it quick.
 316                                 echo $str;
 317                                 continue;
 318                         }
 319
 320                         # We'll have to examine the chunk byte by byte to ensure
 321                         # that it consists of valid UTF-8 sequences, and to see
 322                         # if any of them might not be normalized.
 323                         #
 324                         # Since PHP is not the fastest language on earth, some of
 325                         # this code is a little ugly with inner loop optimizations.
 326
 327                         $len = strlen( $str ) + 1;
 328                         $tail = 0;
 329                         $head = '';
 330                         for( $i = 0; --$len; ++$i ) {
 331                                 if( $tail ) {
 332                                         if( ( $c = $str{$i} ) >= "\x80" && $c < "\xc0" ) {
 333                                                 $sequence .= $c;
 334                                                 if( --$remaining ) {
 335                                                         # Keep adding bytes...
 336                                                         continue;
 337                                                 }
 338
 339                                                 # We have come to the end of the sequence...
 340                                                 $tail = 0;
 341
 342                                                 if( isset( $checkit[$head] ) ) {
 343                                                         # Do some more detailed validity checks, for
 344                                                         # invalid characters and illegal sequences.
 345                                                         if( $head == "\xed" ) {
 346                                                                 # 0xed is relatively frequent in Korean, which
 347                                                                 # abuts the surrogate area, so we're doing
 348                                                                 # this check separately.
 349                                                                 if( $sequence >= UTF8_SURROGATE_FIRST ) {
 350                                                                         echo UTF8_REPLACEMENT;
 351                                                                         continue;
 352                                                                 }
 353                                                         } else {
 354                                                                 $n = ord( $head );
 355                                                                 if(    ($n  < 0xc2 && $sequence <= UTF8_OVERLONG_A)
 356                                                                         || ($n == 0xe0 && $sequence <= UTF8_OVERLONG_B)
 357                                                                         || ($n == 0xef &&
 358                                                                                 ($sequence >= UTF8_FDD0 && $sequence <= UTF8_FDEF)
 359                                                                                 || ($sequence == UTF8_FFFE)
 360                                                                                 || ($sequence == UTF8_FFFF) )
 361                                                                         || ($n == 0xf0 && $sequence <= UTF8_OVERLONG_C)
 362                                                                         || ($n >= 0xf0 && $sequence > UTF8_MAX) ) {
 363                                                                         echo UTF8_REPLACEMENT;
 364                                                                         continue;
 365                                                                 }
 366                                                         }
 367                                                 }
 368
 369                                                 if( isset( $utfCheckOrCombining[$sequence] ) ) {
 370                                                         # If it's NO or MAYBE, we'll have to rip
 371                                                         # the string apart and put it back together.
 372                                                         # That's going to be mighty slow.
 373                                                         $looksNormal = false;
 374                                                 }
 375
 376                                                 # The sequence is legal!
 377                                                 echo $sequence;
 378                                                 $head = '';
 379                                                 continue;
 380                                         }
 381                                         # Not a valid tail byte! DIscard the char we've been building.
 382                                         $tail = false;
 383                                         echo UTF8_REPLACEMENT;
 384                                 }
 385                                 if( $remaining = $tailBytes[$c = $str{$i}] ) {
 386                                         $tail = 1;
 387                                         $sequence = $head = $c;
 388                                 } elseif( $c < "\x80" ) {
 389                                         echo $c;
 390                                 } elseif( $c < "\xc0" ) {
 391                                         # illegal tail bytes or head byte of overlong sequence
 392                                         if( $head == '' ) {
 393                                                 # Don't add if we're continuing a too-long sequence
 394                                                 echo UTF8_REPLACEMENT;
 395                                         }
 396                                 } else {
 397                                         echo UTF8_REPLACEMENT;
 398                                 }
 399                         }
 400                         if( $tail ) {
 401                                 # We ended the chunk in the middle of a sequence;
 402                                 # that's so not cool.
 403                                 echo UTF8_REPLACEMENT;
 404                         }
 405                 }
 406                 $string = ob_get_contents();
 407                 ob_end_clean();
 408                 return $looksNormal;
 409         }
 410
 411         # These take a string and run the normalization on them, without
 412         # checking for validity or any optimization etc. Input must be
 413         # VALID UTF-8!
 414         /**
 415          * @param string $string
 416          * @return string
 417          * @access private
 418          */
 419         function NFC( $string ) {
 420                 return UtfNormal::fastCompose( UtfNormal::NFD( $string ) );
 421         }
 422
 423         /**
 424          * @param string $string
 425          * @return string
 426          * @access private
 427          */
 428         function NFD( $string ) {
 429                 UtfNormal::loadData();
 430                 global $utfCanonicalDecomp;
 431                 return UtfNormal::fastCombiningSort(
 432                         UtfNormal::fastDecompose( $string, $utfCanonicalDecomp ) );
 433         }
 434
 435         /**
 436          * @param string $string
 437          * @return string
 438          * @access private
 439          */
 440         function NFKC( $string ) {
 441                 return UtfNormal::fastCompose( UtfNormal::NFKD( $string ) );
 442         }
 443
 444         /**
 445          * @param string $string
 446          * @return string
 447          * @access private
 448          */
 449         function NFKD( $string ) {
 450                 global $utfCompatibilityDecomp;
 451                 if( !isset( $utfCompatibilityDecomp ) ) {
 452                         require_once( 'UtfNormalDataK.inc' );
 453                 }
 454                 return UtfNormal::fastCombiningSort(
 455                         UtfNormal::fastDecompose( $string, $utfCompatibilityDecomp ) );
 456         }
 457
 458
 459         /**
 460          * Perform decomposition of a UTF-8 string into either D or KD form
 461          * (depending on which decomposition map is passed to us).
 462          * Input is assumed to be *valid* UTF-8. Invalid code will break.
 463          * @access private
 464          * @param string $string Valid UTF-8 string
 465          * @param array $map hash of expanded decomposition map
 466          * @return string a UTF-8 string decomposed, not yet normalized (needs sorting)
 467          */
 468         function fastDecompose( $string, &$map ) {
 469                 UtfNormal::loadData();
 470                 $len = strlen( $string );
 471                 $out = '';
 472                 for( $i = 0; $i < $len; $i++ ) {
 473                         $c = $string{$i};
 474                         $n = ord( $c );
 475                         if( $n < 0x80 ) {
 476                                 # ASCII chars never decompose
 477                                 # THEY ARE IMMORTAL
 478                                 $out .= $c;
 479                                 continue;
 480                         } elseif( $n >= 0xf0 ) {
 481                                 $c = substr( $string, $i, 4 );
 482                                 $i += 3;
 483                         } elseif( $n >= 0xe0 ) {
 484                                 $c = substr( $string, $i, 3 );
 485                                 $i += 2;
 486                         } elseif( $n >= 0xc0 ) {
 487                                 $c = substr( $string, $i, 2 );
 488                                 $i++;
 489                         }
 490                         if( isset( $map[$c] ) ) {
 491                                 $out .= $map[$c];
 492                                 continue;
 493                         } else {
 494                                 if( $c >= UTF8_HANGUL_FIRST && $c <= UTF8_HANGUL_LAST ) {
 495                                         # Decompose a hangul syllable into jamo;
 496                                         # hardcoded for three-byte UTF-8 sequence.
 497                                         # A lookup table would be slightly faster,
 498                                         # but adds a lot of memory & disk needs.
 499                                         #
 500                                         $index = ( (ord( $c{0} ) & 0x0f) << 12
 501                                                  | (ord( $c{1} ) & 0x3f) <<  6
 502                                                  | (ord( $c{2} ) & 0x3f) )
 503                                                - UNICODE_HANGUL_FIRST;
 504                                         $l = IntVal( $index / UNICODE_HANGUL_NCOUNT );
 505                                         $v = IntVal( ($index % UNICODE_HANGUL_NCOUNT) / UNICODE_HANGUL_TCOUNT);
 506                                         $t = $index % UNICODE_HANGUL_TCOUNT;
 507                                         $out .= "\xe1\x84" . chr( 0x80 + $l ) . "\xe1\x85" . chr( 0xa1 + $v );
 508                                         if( $t >= 25 ) {
 509                                                 $out .= "\xe1\x87" . chr( 0x80 + $t - 25 );
 510                                         } elseif( $t ) {
 511                                                 $out .= "\xe1\x86" . chr( 0xa7 + $t );
 512                                         }
 513                                         continue;
 514                                 }
 515                         }
 516                         $out .= $c;
 517                 }
 518                 return $out;
 519         }
 520
 521         /**
 522          * Sorts combining characters into canonical order. This is the
 523          * final step in creating decomposed normal forms D and KD.
 524          * @access private
 525          * @param string $string a valid, decomposed UTF-8 string. Input is not validated.
 526          * @return string a UTF-8 string with combining characters sorted in canonical order
 527          */
 528         function fastCombiningSort( $string ) {
 529                 UtfNormal::loadData();
 530                 global $utfCombiningClass;
 531                 $len = strlen( $string );
 532                 $out = '';
 533                 $combiners = array();
 534                 $lastClass = -1;
 535                 for( $i = 0; $i < $len; $i++ ) {
 536                         $c = $string{$i};
 537                         $n = ord( $c );
 538                         if( $n >= 0x80 ) {
 539                                 if( $n >= 0xf0 ) {
 540                                         $c = substr( $string, $i, 4 );
 541                                         $i += 3;
 542                                 } elseif( $n >= 0xe0 ) {
 543                                         $c = substr( $string, $i, 3 );
 544                                         $i += 2;
 545                                 } elseif( $n >= 0xc0 ) {
 546                                         $c = substr( $string, $i, 2 );
 547                                         $i++;
 548                                 }
 549                                 if( isset( $utfCombiningClass[$c] ) ) {
 550                                         $lastClass = $utfCombiningClass[$c];
 551                                         @$combiners[$lastClass] .= $c;
 552                                         continue;
 553                                 }
 554                         }
 555                         if( $lastClass ) {
 556                                 ksort( $combiners );
 557                                 $out .= implode( '', $combiners );
 558                                 $combiners = array();
 559                         }
 560                         $out .= $c;
 561                         $lastClass = 0;
 562                 }
 563                 if( $lastClass ) {
 564                         ksort( $combiners );
 565                         $out .= implode( '', $combiners );
 566                 }
 567                 return $out;
 568         }
 569
 570         /**
 571          * Produces canonically composed sequences, i.e. normal form C or KC.
 572          *
 573          * @access private
 574          * @param string $string a valid UTF-8 string in sorted normal form D or KD. Input is not validated.
 575          * @return string a UTF-8 string with canonical precomposed characters used where possible
 576          */
 577         function fastCompose( $string ) {
 578                 UtfNormal::loadData();
 579                 global $utfCanonicalComp, $utfCombiningClass;
 580                 $len = strlen( $string );
 581                 $out = '';
 582                 $lastClass = -1;
 583                 $startChar = '';
 584                 $combining = '';
 585                 $x1 = ord(substr(UTF8_HANGUL_VBASE,0,1));
 586                 $x2 = ord(substr(UTF8_HANGUL_TEND,0,1));
 587                 for( $i = 0; $i < $len; $i++ ) {
 588                         $c = $string{$i};
 589                         $n = ord( $c );
 590                         if( $n < 0x80 ) {
 591                                 # No combining characters here...
 592                                 $out .= $startChar;
 593                                 $out .= $combining;
 594                                 $startChar = $c;
 595                                 $combining = '';
 596                                 $lastClass = 0;
 597                                 continue;
 598                         } elseif( $n >= 0xf0 ) {
 599                                 $c = substr( $string, $i, 4 );
 600                                 $i += 3;
 601                         } elseif( $n >= 0xe0 ) {
 602                                 $c = substr( $string, $i, 3 );
 603                                 $i += 2;
 604                         } elseif( $n >= 0xc0 ) {
 605                                 $c = substr( $string, $i, 2 );
 606                                 $i++;
 607                         }
 608                         $pair = $startChar . $c;
 609                         if( $n > 0x80 ) {
 610                                 if( isset( $utfCombiningClass[$c] ) ) {
 611                                         # A combining char; see what we can do with it
 612                                         $class = $utfCombiningClass[$c];
 613                                         if( !empty( $startChar ) &&
 614                                                 $lastClass < $class &&
 615                                                 $class > 0 &&
 616                                                 isset( $utfCanonicalComp[$pair] ) ) {
 617                                                 $startChar = $utfCanonicalComp[$pair];
 618                                                 $class = 0;
 619                                         } else {
 620                                                 $combining .= $c;
 621                                         }
 622                                         $lastClass = $class;
 623                                         continue;
 624                                 }
 625                         }
 626                         # New start char
 627                         if( $lastClass == 0 ) {
 628                                 if( isset( $utfCanonicalComp[$pair] ) ) {
 629                                         $startChar = $utfCanonicalComp[$pair];
 630                                         continue;
 631                                 }
 632                                 if( $n >= $x1 && $n <= $x2 ) {
 633                                         # WARNING: Hangul code is painfully slow.
 634                                         # I apologize for this ugly, ugly code; however
 635                                         # performance is even more teh suck if we call
 636                                         # out to nice clean functions. Lookup tables are
 637                                         # marginally faster, but require a lot of space.
 638                                         #
 639                                         if( $c >= UTF8_HANGUL_VBASE &&
 640                                                 $c <= UTF8_HANGUL_VEND &&
 641                                                 $startChar >= UTF8_HANGUL_LBASE &&
 642                                                 $startChar <= UTF8_HANGUL_LEND ) {
 643                                                 #
 644                                                 #$lIndex = utf8ToCodepoint( $startChar ) - UNICODE_HANGUL_LBASE;
 645                                                 #$vIndex = utf8ToCodepoint( $c ) - UNICODE_HANGUL_VBASE;
 646                                                 $lIndex = ord( $startChar{2} ) - 0x80;
 647                                                 $vIndex = ord( $c{2}         ) - 0xa1;
 648
 649                                                 $hangulPoint = UNICODE_HANGUL_FIRST +
 650                                                         UNICODE_HANGUL_TCOUNT *
 651                                                         (UNICODE_HANGUL_VCOUNT * $lIndex + $vIndex);
 652
 653                                                 # Hardcode the limited-range UTF-8 conversion:
 654                                                 $startChar = chr( $hangulPoint >> 12 & 0x0f | 0xe0 ) .
 655                                                                          chr( $hangulPoint >>  6 & 0x3f | 0x80 ) .
 656                                                                          chr( $hangulPoint       & 0x3f | 0x80 );
 657                                                 continue;
 658                                         } elseif( $c >= UTF8_HANGUL_TBASE &&
 659                                                           $c <= UTF8_HANGUL_TEND &&
 660                                                           $startChar >= UTF8_HANGUL_FIRST &&
 661                                                           $startChar <= UTF8_HANGUL_LAST ) {
 662                                                 # $tIndex = utf8ToCodepoint( $c ) - UNICODE_HANGUL_TBASE;
 663                                                 $tIndex = ord( $c{2} ) - 0xa7;
 664                                                 if( $tIndex < 0 ) $tIndex = ord( $c{2} ) - 0x80 + (0x11c0 - 0x11a7);
 665
 666                                                 # Increment the code point by $tIndex, without
 667                                                 # the function overhead of decoding and recoding UTF-8
 668                                                 #
 669                                                 $tail = ord( $startChar{2} ) + $tIndex;
 670                                                 if( $tail > 0xbf ) {
 671                                                         $tail -= 0x40;
 672                                                         $mid = ord( $startChar{1} ) + 1;
 673                                                         if( $mid > 0xbf ) {
 674                                                                 $startChar{0} = chr( ord( $startChar{0} ) + 1 );
 675                                                                 $mid -= 0x40;
 676                                                         }
 677                                                         $startChar{1} = chr( $mid );
 678                                                 }
 679                                                 $startChar{2} = chr( $tail );
 680                                                 continue;
 681                                         }
 682                                 }
 683                         }
 684                         $out .= $startChar;
 685                         $out .= $combining;
 686                         $startChar = $c;
 687                         $combining = '';
 688                         $lastClass = 0;
 689                 }
 690                 $out .= $startChar . $combining;
 691                 return $out;
 692         }
 693
 694         /**
 695          * This is just used for the benchmark, comparing how long it takes to
 696          * interate through a string without really doing anything of substance.
 697          * @param string $string
 698          * @return string
 699          */
 700         function placebo( $string ) {
 701                 $len = strlen( $string );
 702                 $out = '';
 703                 for( $i = 0; $i < $len; $i++ ) {
 704                         $out .= $string{$i};
 705                 }
 706                 return $out;
 707         }
 708 }
 709
 710 ?>