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                 preg_replace( '/[\x00-\x08\x0b\x0c\x0e-\x1f]/', UTF8_REPLACEMENT, $string );
 262
 263                 # ASCII is always valid NFC!
 264                 if( !preg_match( '/[\x80-\xff]/', $string ) ) return true;
 265
 266                 UtfNormal::loadData();
 267                 global $utfCheckNFC, $utfCombiningClass;
 268
 269                 static $checkit = null, $tailBytes = null;
 270                 if( !isset( $checkit ) ) {
 271                         # Head bytes for sequences which we should do further validity checks
 272                         $checkit = array_flip(
 273                                         array( 0xc0, 0xc1, 0xe0, 0xed, 0xef,
 274                                                    0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7,
 275                                                    0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff ) );
 276
 277                         $tailBytes = array();
 278                         for( $n = 0xc0; $n < 0xfe; $n++ ) {
 279                                 if( $n < 0xe0 ) {
 280                                         $remaining = 1;
 281                                 } elseif( $n < 0xf0 ) {
 282                                         $remaining = 2;
 283                                 } elseif( $n < 0xf8 ) {
 284                                         $remaining = 3;
 285                                 } elseif( $n < 0xfc ) {
 286                                         $remaining = 4;
 287                                 } elseif( $n < 0xfe ) {
 288                                         $remaining = 5;
 289                                 }
 290                                 $tailBytes[$n] = $remaining;
 291                         }
 292                 }
 293
 294                 $len = strlen( $string );
 295                 $out = '';
 296                 $tail = false;
 297                 $looksNormal = true;
 298                 $head = 0;
 299
 300                 for( $i = 0; $i < $len; $i++ ) {
 301                         $c = $string{$i};
 302                         $n = ord( $c );
 303                         if( $tail ) {
 304                                 if( $n >= 0x80 && $n < 0xc0 ) {
 305                                         $sequence .= $c;
 306                                         if( --$remaining ) {
 307                                                 # Keep adding bytes...
 308                                                 continue;
 309                                         }
 310
 311                                         if( isset( $checkit[$head] ) ) {
 312                                                 # Do some more detailed validity checks, for
 313                                                 # invalid characters and illegal sequences.
 314                                                 if( ( $head == 0xed && $sequence >= UTF8_SURROGATE_FIRST
 315                                                                 && $sequence <= UTF8_SURROGATE_LAST)
 316                                                         || ($head  < 0xc2 && $sequence <= UTF8_OVERLONG_A)
 317                                                         || ($head == 0xe0 && $sequence <= UTF8_OVERLONG_B)
 318                                                         || ($head == 0xef &&
 319                                                                 ($sequence >= UTF8_FDD0 && $sequence <= UTF8_FDEF)
 320                                                                 || ($sequence == UTF8_FFFE)
 321                                                                 || ($sequence == UTF8_FFFF) )
 322                                                         || ($head == 0xf0 && $sequence <= UTF8_OVERLONG_C)
 323                                                         || ($head >= 0xf0 && $sequence > UTF8_MAX) ) {
 324                                                         $out .= UTF8_REPLACEMENT;
 325                                                         $tail = false;
 326                                                         continue;
 327                                                 }
 328                                         }
 329
 330                                         if( isset( $utfCheckNFC[$sequence] ) ||
 331                                                 isset( $utfCombiningClass[$sequence] ) ) {
 332                                                 # If it's NO or MAYBE, we'll have to do the slow check.
 333                                                 $looksNormal = false;
 334                                         }
 335
 336                                         # The sequence is legal!
 337                                         $out .= $sequence;
 338                                         $tail = false;
 339                                         $head = 0;
 340                                         continue;
 341                                 }
 342                                 # Not a valid tail byte! DIscard the char we've been building.
 343                                 #printf ("Invalid '%x' in tail with %d remaining bytes\n", $n, $remaining );
 344                                 $tail = false;
 345                                 $out .= UTF8_REPLACEMENT;
 346                         }
 347                         if( $n < 0x80 ) {
 348                                 # Friendly ASCII chars.
 349                                 $out .= $c;
 350                         } elseif( $n < 0xc0 ) {
 351                                 # illegal tail bytes or head byte of overlong sequence
 352                                 if( $head == 0 ) {
 353                                         # Don't add if we're continuing a too-long sequence
 354                                         $out .= UTF8_REPLACEMENT;
 355                                 }
 356                         } elseif( $n < 0xfe ) {
 357                                 $tail = true;
 358                                 $remaining = $tailBytes[$n];
 359                                 $sequence = $c;
 360                                 $head = $n;
 361                         } else {
 362                                 $out .= UTF8_REPLACEMENT;
 363                         }
 364                 }
 365                 if( $tail ) {
 366                         $out .= UTF8_REPLACEMENT;
 367                 }
 368                 $string = $out;
 369                 return $looksNormal;
 370         }
 371
 372         # These take a string and run the normalization on them, without
 373         # checking for validity or any optimization etc. Input must be
 374         # VALID UTF-8!
 375         /**
 376          * @param string $string
 377          * @return string
 378          * @access private
 379          */
 380         function NFC( $string ) {
 381                 return UtfNormal::fastCompose( UtfNormal::NFD( $string ) );
 382         }
 383
 384         /**
 385          * @param string $string
 386          * @return string
 387          * @access private
 388          */
 389         function NFD( $string ) {
 390                 UtfNormal::loadData();
 391                 global $utfCanonicalDecomp;
 392                 return UtfNormal::fastCombiningSort(
 393                         UtfNormal::fastDecompose( $string, $utfCanonicalDecomp ) );
 394         }
 395
 396         /**
 397          * @param string $string
 398          * @return string
 399          * @access private
 400          */
 401         function NFKC( $string ) {
 402                 return UtfNormal::fastCompose( UtfNormal::NFKD( $string ) );
 403         }
 404
 405         /**
 406          * @param string $string
 407          * @return string
 408          * @access private
 409          */
 410         function NFKD( $string ) {
 411                 global $utfCompatibilityDecomp;
 412                 if( !isset( $utfCompatibilityDecomp ) ) {
 413                         require_once( 'UtfNormalDataK.inc' );
 414                 }
 415                 return UtfNormal::fastCombiningSort(
 416                         UtfNormal::fastDecompose( $string, $utfCompatibilityDecomp ) );
 417         }
 418
 419
 420         /**
 421          * Perform decomposition of a UTF-8 string into either D or KD form
 422          * (depending on which decomposition map is passed to us).
 423          * Input is assumed to be *valid* UTF-8. Invalid code will break.
 424          * @access private
 425          * @param string $string Valid UTF-8 string
 426          * @param array $map hash of expanded decomposition map
 427          * @return string a UTF-8 string decomposed, not yet normalized (needs sorting)
 428          */
 429         function fastDecompose( $string, &$map ) {
 430                 UtfNormal::loadData();
 431                 $len = strlen( $string );
 432                 $out = '';
 433                 for( $i = 0; $i < $len; $i++ ) {
 434                         $c = $string{$i};
 435                         $n = ord( $c );
 436                         if( $n < 0x80 ) {
 437                                 # ASCII chars never decompose
 438                                 # THEY ARE IMMORTAL
 439                                 $out .= $c;
 440                                 continue;
 441                         } elseif( $n >= 0xf0 ) {
 442                                 $c = substr( $string, $i, 4 );
 443                                 $i += 3;
 444                         } elseif( $n >= 0xe0 ) {
 445                                 $c = substr( $string, $i, 3 );
 446                                 $i += 2;
 447                         } elseif( $n >= 0xc0 ) {
 448                                 $c = substr( $string, $i, 2 );
 449                                 $i++;
 450                         }
 451                         if( isset( $map[$c] ) ) {
 452                                 $out .= $map[$c];
 453                                 continue;
 454                         } else {
 455                                 if( $c >= UTF8_HANGUL_FIRST && $c <= UTF8_HANGUL_LAST ) {
 456                                         # Decompose a hangul syllable into jamo;
 457                                         # hardcoded for three-byte UTF-8 sequence.
 458                                         # A lookup table would be slightly faster,
 459                                         # but adds a lot of memory & disk needs.
 460                                         #
 461                                         $index = ( (ord( $c{0} ) & 0x0f) << 12
 462                                                  | (ord( $c{1} ) & 0x3f) <<  6
 463                                                  | (ord( $c{2} ) & 0x3f) )
 464                                                - UNICODE_HANGUL_FIRST;
 465                                         $l = IntVal( $index / UNICODE_HANGUL_NCOUNT );
 466                                         $v = IntVal( ($index % UNICODE_HANGUL_NCOUNT) / UNICODE_HANGUL_TCOUNT);
 467                                         $t = $index % UNICODE_HANGUL_TCOUNT;
 468                                         $out .= "\xe1\x84" . chr( 0x80 + $l ) . "\xe1\x85" . chr( 0xa1 + $v );
 469                                         if( $t >= 25 ) {
 470                                                 $out .= "\xe1\x87" . chr( 0x80 + $t - 25 );
 471                                         } elseif( $t ) {
 472                                                 $out .= "\xe1\x86" . chr( 0xa7 + $t );
 473                                         }
 474                                         continue;
 475                                 }
 476                         }
 477                         $out .= $c;
 478                 }
 479                 return $out;
 480         }
 481
 482         /**
 483          * Sorts combining characters into canonical order. This is the
 484          * final step in creating decomposed normal forms D and KD.
 485          * @access private
 486          * @param string $string a valid, decomposed UTF-8 string. Input is not validated.
 487          * @return string a UTF-8 string with combining characters sorted in canonical order
 488          */
 489         function fastCombiningSort( $string ) {
 490                 UtfNormal::loadData();
 491                 global $utfCombiningClass;
 492                 $len = strlen( $string );
 493                 $out = '';
 494                 $combiners = array();
 495                 $lastClass = -1;
 496                 for( $i = 0; $i < $len; $i++ ) {
 497                         $c = $string{$i};
 498                         $n = ord( $c );
 499                         if( $n >= 0x80 ) {
 500                                 if( $n >= 0xf0 ) {
 501                                         $c = substr( $string, $i, 4 );
 502                                         $i += 3;
 503                                 } elseif( $n >= 0xe0 ) {
 504                                         $c = substr( $string, $i, 3 );
 505                                         $i += 2;
 506                                 } elseif( $n >= 0xc0 ) {
 507                                         $c = substr( $string, $i, 2 );
 508                                         $i++;
 509                                 }
 510                                 if( isset( $utfCombiningClass[$c] ) ) {
 511                                         $lastClass = $utfCombiningClass[$c];
 512                                         @$combiners[$lastClass] .= $c;
 513                                         continue;
 514                                 }
 515                         }
 516                         if( $lastClass ) {
 517                                 ksort( $combiners );
 518                                 $out .= implode( '', $combiners );
 519                                 $combiners = array();
 520                         }
 521                         $out .= $c;
 522                         $lastClass = 0;
 523                 }
 524                 if( $lastClass ) {
 525                         ksort( $combiners );
 526                         $out .= implode( '', $combiners );
 527                 }
 528                 return $out;
 529         }
 530
 531         /**
 532          * Produces canonically composed sequences, i.e. normal form C or KC.
 533          *
 534          * @access private
 535          * @param string $string a valid UTF-8 string in sorted normal form D or KD. Input is not validated.
 536          * @return string a UTF-8 string with canonical precomposed characters used where possible
 537          */
 538         function fastCompose( $string ) {
 539                 UtfNormal::loadData();
 540                 global $utfCanonicalComp, $utfCombiningClass;
 541                 $len = strlen( $string );
 542                 $out = '';
 543                 $lastClass = -1;
 544                 $startChar = '';
 545                 $combining = '';
 546                 $x1 = ord(substr(UTF8_HANGUL_VBASE,0,1));
 547                 $x2 = ord(substr(UTF8_HANGUL_TEND,0,1));
 548                 for( $i = 0; $i < $len; $i++ ) {
 549                         $c = $string{$i};
 550                         $n = ord( $c );
 551                         if( $n < 0x80 ) {
 552                                 # No combining characters here...
 553                                 $out .= $startChar;
 554                                 $out .= $combining;
 555                                 $startChar = $c;
 556                                 $combining = '';
 557                                 $lastClass = 0;
 558                                 continue;
 559                         } elseif( $n >= 0xf0 ) {
 560                                 $c = substr( $string, $i, 4 );
 561                                 $i += 3;
 562                         } elseif( $n >= 0xe0 ) {
 563                                 $c = substr( $string, $i, 3 );
 564                                 $i += 2;
 565                         } elseif( $n >= 0xc0 ) {
 566                                 $c = substr( $string, $i, 2 );
 567                                 $i++;
 568                         }
 569                         $pair = $startChar . $c;
 570                         if( $n > 0x80 ) {
 571                                 if( isset( $utfCombiningClass[$c] ) ) {
 572                                         # A combining char; see what we can do with it
 573                                         $class = $utfCombiningClass[$c];
 574                                         if( !empty( $startChar ) &&
 575                                                 $lastClass < $class &&
 576                                                 $class > 0 &&
 577                                                 isset( $utfCanonicalComp[$pair] ) ) {
 578                                                 $startChar = $utfCanonicalComp[$pair];
 579                                                 $class = 0;
 580                                         } else {
 581                                                 $combining .= $c;
 582                                         }
 583                                         $lastClass = $class;
 584                                         continue;
 585                                 }
 586                         }
 587                         # New start char
 588                         if( $lastClass == 0 ) {
 589                                 if( isset( $utfCanonicalComp[$pair] ) ) {
 590                                         $startChar = $utfCanonicalComp[$pair];
 591                                         continue;
 592                                 }
 593                                 if( $n >= $x1 && $n <= $x2 ) {
 594                                         # WARNING: Hangul code is painfully slow.
 595                                         # I apologize for this ugly, ugly code; however
 596                                         # performance is even more teh suck if we call
 597                                         # out to nice clean functions. Lookup tables are
 598                                         # marginally faster, but require a lot of space.
 599                                         #
 600                                         if( $c >= UTF8_HANGUL_VBASE &&
 601                                                 $c <= UTF8_HANGUL_VEND &&
 602                                                 $startChar >= UTF8_HANGUL_LBASE &&
 603                                                 $startChar <= UTF8_HANGUL_LEND ) {
 604                                                 #
 605                                                 #$lIndex = utf8ToCodepoint( $startChar ) - UNICODE_HANGUL_LBASE;
 606                                                 #$vIndex = utf8ToCodepoint( $c ) - UNICODE_HANGUL_VBASE;
 607                                                 $lIndex = ord( $startChar{2} ) - 0x80;
 608                                                 $vIndex = ord( $c{2}         ) - 0xa1;
 609
 610                                                 $hangulPoint = UNICODE_HANGUL_FIRST +
 611                                                         UNICODE_HANGUL_TCOUNT *
 612                                                         (UNICODE_HANGUL_VCOUNT * $lIndex + $vIndex);
 613
 614                                                 # Hardcode the limited-range UTF-8 conversion:
 615                                                 $startChar = chr( $hangulPoint >> 12 & 0x0f | 0xe0 ) .
 616                                                                          chr( $hangulPoint >>  6 & 0x3f | 0x80 ) .
 617                                                                          chr( $hangulPoint       & 0x3f | 0x80 );
 618                                                 continue;
 619                                         } elseif( $c >= UTF8_HANGUL_TBASE &&
 620                                                           $c <= UTF8_HANGUL_TEND &&
 621                                                           $startChar >= UTF8_HANGUL_FIRST &&
 622                                                           $startChar <= UTF8_HANGUL_LAST ) {
 623                                                 # $tIndex = utf8ToCodepoint( $c ) - UNICODE_HANGUL_TBASE;
 624                                                 $tIndex = ord( $c{2} ) - 0xa7;
 625                                                 if( $tIndex < 0 ) $tIndex = ord( $c{2} ) - 0x80 + (0x11c0 - 0x11a7);
 626
 627                                                 # Increment the code point by $tIndex, without
 628                                                 # the function overhead of decoding and recoding UTF-8
 629                                                 #
 630                                                 $tail = ord( $startChar{2} ) + $tIndex;
 631                                                 if( $tail > 0xbf ) {
 632                                                         $tail -= 0x40;
 633                                                         $mid = ord( $startChar{1} ) + 1;
 634                                                         if( $mid > 0xbf ) {
 635                                                                 $startChar{0} = chr( ord( $startChar{0} ) + 1 );
 636                                                                 $mid -= 0x40;
 637                                                         }
 638                                                         $startChar{1} = chr( $mid );
 639                                                 }
 640                                                 $startChar{2} = chr( $tail );
 641                                                 continue;
 642                                         }
 643                                 }
 644                         }
 645                         $out .= $startChar;
 646                         $out .= $combining;
 647                         $startChar = $c;
 648                         $combining = '';
 649                         $lastClass = 0;
 650                 }
 651                 $out .= $startChar . $combining;
 652                 return $out;
 653         }
 654
 655         /**
 656          * This is just used for the benchmark, comparing how long it takes to
 657          * interate through a string without really doing anything of substance.
 658          * @param string $string
 659          * @return string
 660          */
 661         function placebo( $string ) {
 662                 $len = strlen( $string );
 663                 $out = '';
 664                 for( $i = 0; $i < $len; $i++ ) {
 665                         $out .= $string{$i};
 666                 }
 667                 return $out;
 668         }
 669 }
 670
 671 ?>