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                                 # We can speed things up a bit for latin-based scripts
 350                                 # where they tend to come in groups:
 351                                 $out .= $c;
 352                                 $i++;
 353                                 while( $i < $len && ( $c = $string{$i} ) < "\x80" ) {
 354                                         $out .= $c;
 355                                         $i++;
 356                                 }
 357                                 $i--;
 358                         } elseif( $n < 0xc0 ) {
 359                                 # illegal tail bytes or head byte of overlong sequence
 360                                 if( $head == 0 ) {
 361                                         # Don't add if we're continuing a too-long sequence
 362                                         $out .= UTF8_REPLACEMENT;
 363                                 }
 364                         } elseif( $n < 0xfe ) {
 365                                 $tail = true;
 366                                 $remaining = $tailBytes[$n];
 367                                 $sequence = $c;
 368                                 $head = $n;
 369                         } else {
 370                                 $out .= UTF8_REPLACEMENT;
 371                         }
 372                 }
 373                 if( $tail ) {
 374                         $out .= UTF8_REPLACEMENT;
 375                 }
 376                 $string = $out;
 377                 return $looksNormal;
 378         }
 379
 380         # These take a string and run the normalization on them, without
 381         # checking for validity or any optimization etc. Input must be
 382         # VALID UTF-8!
 383         /**
 384          * @param string $string
 385          * @return string
 386          * @access private
 387          */
 388         function NFC( $string ) {
 389                 return UtfNormal::fastCompose( UtfNormal::NFD( $string ) );
 390         }
 391
 392         /**
 393          * @param string $string
 394          * @return string
 395          * @access private
 396          */
 397         function NFD( $string ) {
 398                 UtfNormal::loadData();
 399                 global $utfCanonicalDecomp;
 400                 return UtfNormal::fastCombiningSort(
 401                         UtfNormal::fastDecompose( $string, $utfCanonicalDecomp ) );
 402         }
 403
 404         /**
 405          * @param string $string
 406          * @return string
 407          * @access private
 408          */
 409         function NFKC( $string ) {
 410                 return UtfNormal::fastCompose( UtfNormal::NFKD( $string ) );
 411         }
 412
 413         /**
 414          * @param string $string
 415          * @return string
 416          * @access private
 417          */
 418         function NFKD( $string ) {
 419                 global $utfCompatibilityDecomp;
 420                 if( !isset( $utfCompatibilityDecomp ) ) {
 421                         require_once( 'UtfNormalDataK.inc' );
 422                 }
 423                 return UtfNormal::fastCombiningSort(
 424                         UtfNormal::fastDecompose( $string, $utfCompatibilityDecomp ) );
 425         }
 426
 427
 428         /**
 429          * Perform decomposition of a UTF-8 string into either D or KD form
 430          * (depending on which decomposition map is passed to us).
 431          * Input is assumed to be *valid* UTF-8. Invalid code will break.
 432          * @access private
 433          * @param string $string Valid UTF-8 string
 434          * @param array $map hash of expanded decomposition map
 435          * @return string a UTF-8 string decomposed, not yet normalized (needs sorting)
 436          */
 437         function fastDecompose( $string, &$map ) {
 438                 UtfNormal::loadData();
 439                 $len = strlen( $string );
 440                 $out = '';
 441                 for( $i = 0; $i < $len; $i++ ) {
 442                         $c = $string{$i};
 443                         $n = ord( $c );
 444                         if( $n < 0x80 ) {
 445                                 # ASCII chars never decompose
 446                                 # THEY ARE IMMORTAL
 447                                 $out .= $c;
 448                                 continue;
 449                         } elseif( $n >= 0xf0 ) {
 450                                 $c = substr( $string, $i, 4 );
 451                                 $i += 3;
 452                         } elseif( $n >= 0xe0 ) {
 453                                 $c = substr( $string, $i, 3 );
 454                                 $i += 2;
 455                         } elseif( $n >= 0xc0 ) {
 456                                 $c = substr( $string, $i, 2 );
 457                                 $i++;
 458                         }
 459                         if( isset( $map[$c] ) ) {
 460                                 $out .= $map[$c];
 461                                 continue;
 462                         } else {
 463                                 if( $c >= UTF8_HANGUL_FIRST && $c <= UTF8_HANGUL_LAST ) {
 464                                         # Decompose a hangul syllable into jamo;
 465                                         # hardcoded for three-byte UTF-8 sequence.
 466                                         # A lookup table would be slightly faster,
 467                                         # but adds a lot of memory & disk needs.
 468                                         #
 469                                         $index = ( (ord( $c{0} ) & 0x0f) << 12
 470                                                  | (ord( $c{1} ) & 0x3f) <<  6
 471                                                  | (ord( $c{2} ) & 0x3f) )
 472                                                - UNICODE_HANGUL_FIRST;
 473                                         $l = IntVal( $index / UNICODE_HANGUL_NCOUNT );
 474                                         $v = IntVal( ($index % UNICODE_HANGUL_NCOUNT) / UNICODE_HANGUL_TCOUNT);
 475                                         $t = $index % UNICODE_HANGUL_TCOUNT;
 476                                         $out .= "\xe1\x84" . chr( 0x80 + $l ) . "\xe1\x85" . chr( 0xa1 + $v );
 477                                         if( $t >= 25 ) {
 478                                                 $out .= "\xe1\x87" . chr( 0x80 + $t - 25 );
 479                                         } elseif( $t ) {
 480                                                 $out .= "\xe1\x86" . chr( 0xa7 + $t );
 481                                         }
 482                                         continue;
 483                                 }
 484                         }
 485                         $out .= $c;
 486                 }
 487                 return $out;
 488         }
 489
 490         /**
 491          * Sorts combining characters into canonical order. This is the
 492          * final step in creating decomposed normal forms D and KD.
 493          * @access private
 494          * @param string $string a valid, decomposed UTF-8 string. Input is not validated.
 495          * @return string a UTF-8 string with combining characters sorted in canonical order
 496          */
 497         function fastCombiningSort( $string ) {
 498                 UtfNormal::loadData();
 499                 global $utfCombiningClass;
 500                 $len = strlen( $string );
 501                 $out = '';
 502                 $combiners = array();
 503                 $lastClass = -1;
 504                 for( $i = 0; $i < $len; $i++ ) {
 505                         $c = $string{$i};
 506                         $n = ord( $c );
 507                         if( $n >= 0x80 ) {
 508                                 if( $n >= 0xf0 ) {
 509                                         $c = substr( $string, $i, 4 );
 510                                         $i += 3;
 511                                 } elseif( $n >= 0xe0 ) {
 512                                         $c = substr( $string, $i, 3 );
 513                                         $i += 2;
 514                                 } elseif( $n >= 0xc0 ) {
 515                                         $c = substr( $string, $i, 2 );
 516                                         $i++;
 517                                 }
 518                                 if( isset( $utfCombiningClass[$c] ) ) {
 519                                         $lastClass = $utfCombiningClass[$c];
 520                                         @$combiners[$lastClass] .= $c;
 521                                         continue;
 522                                 }
 523                         }
 524                         if( $lastClass ) {
 525                                 ksort( $combiners );
 526                                 $out .= implode( '', $combiners );
 527                                 $combiners = array();
 528                         }
 529                         $out .= $c;
 530                         $lastClass = 0;
 531                 }
 532                 if( $lastClass ) {
 533                         ksort( $combiners );
 534                         $out .= implode( '', $combiners );
 535                 }
 536                 return $out;
 537         }
 538
 539         /**
 540          * Produces canonically composed sequences, i.e. normal form C or KC.
 541          *
 542          * @access private
 543          * @param string $string a valid UTF-8 string in sorted normal form D or KD. Input is not validated.
 544          * @return string a UTF-8 string with canonical precomposed characters used where possible
 545          */
 546         function fastCompose( $string ) {
 547                 UtfNormal::loadData();
 548                 global $utfCanonicalComp, $utfCombiningClass;
 549                 $len = strlen( $string );
 550                 $out = '';
 551                 $lastClass = -1;
 552                 $startChar = '';
 553                 $combining = '';
 554                 $x1 = ord(substr(UTF8_HANGUL_VBASE,0,1));
 555                 $x2 = ord(substr(UTF8_HANGUL_TEND,0,1));
 556                 for( $i = 0; $i < $len; $i++ ) {
 557                         $c = $string{$i};
 558                         $n = ord( $c );
 559                         if( $n < 0x80 ) {
 560                                 # No combining characters here...
 561                                 $out .= $startChar;
 562                                 $out .= $combining;
 563                                 $startChar = $c;
 564                                 $combining = '';
 565                                 $lastClass = 0;
 566                                 continue;
 567                         } elseif( $n >= 0xf0 ) {
 568                                 $c = substr( $string, $i, 4 );
 569                                 $i += 3;
 570                         } elseif( $n >= 0xe0 ) {
 571                                 $c = substr( $string, $i, 3 );
 572                                 $i += 2;
 573                         } elseif( $n >= 0xc0 ) {
 574                                 $c = substr( $string, $i, 2 );
 575                                 $i++;
 576                         }
 577                         $pair = $startChar . $c;
 578                         if( $n > 0x80 ) {
 579                                 if( isset( $utfCombiningClass[$c] ) ) {
 580                                         # A combining char; see what we can do with it
 581                                         $class = $utfCombiningClass[$c];
 582                                         if( !empty( $startChar ) &&
 583                                                 $lastClass < $class &&
 584                                                 $class > 0 &&
 585                                                 isset( $utfCanonicalComp[$pair] ) ) {
 586                                                 $startChar = $utfCanonicalComp[$pair];
 587                                                 $class = 0;
 588                                         } else {
 589                                                 $combining .= $c;
 590                                         }
 591                                         $lastClass = $class;
 592                                         continue;
 593                                 }
 594                         }
 595                         # New start char
 596                         if( $lastClass == 0 ) {
 597                                 if( isset( $utfCanonicalComp[$pair] ) ) {
 598                                         $startChar = $utfCanonicalComp[$pair];
 599                                         continue;
 600                                 }
 601                                 if( $n >= $x1 && $n <= $x2 ) {
 602                                         # WARNING: Hangul code is painfully slow.
 603                                         # I apologize for this ugly, ugly code; however
 604                                         # performance is even more teh suck if we call
 605                                         # out to nice clean functions. Lookup tables are
 606                                         # marginally faster, but require a lot of space.
 607                                         #
 608                                         if( $c >= UTF8_HANGUL_VBASE &&
 609                                                 $c <= UTF8_HANGUL_VEND &&
 610                                                 $startChar >= UTF8_HANGUL_LBASE &&
 611                                                 $startChar <= UTF8_HANGUL_LEND ) {
 612                                                 #
 613                                                 #$lIndex = utf8ToCodepoint( $startChar ) - UNICODE_HANGUL_LBASE;
 614                                                 #$vIndex = utf8ToCodepoint( $c ) - UNICODE_HANGUL_VBASE;
 615                                                 $lIndex = ord( $startChar{2} ) - 0x80;
 616                                                 $vIndex = ord( $c{2}         ) - 0xa1;
 617
 618                                                 $hangulPoint = UNICODE_HANGUL_FIRST +
 619                                                         UNICODE_HANGUL_TCOUNT *
 620                                                         (UNICODE_HANGUL_VCOUNT * $lIndex + $vIndex);
 621
 622                                                 # Hardcode the limited-range UTF-8 conversion:
 623                                                 $startChar = chr( $hangulPoint >> 12 & 0x0f | 0xe0 ) .
 624                                                                          chr( $hangulPoint >>  6 & 0x3f | 0x80 ) .
 625                                                                          chr( $hangulPoint       & 0x3f | 0x80 );
 626                                                 continue;
 627                                         } elseif( $c >= UTF8_HANGUL_TBASE &&
 628                                                           $c <= UTF8_HANGUL_TEND &&
 629                                                           $startChar >= UTF8_HANGUL_FIRST &&
 630                                                           $startChar <= UTF8_HANGUL_LAST ) {
 631                                                 # $tIndex = utf8ToCodepoint( $c ) - UNICODE_HANGUL_TBASE;
 632                                                 $tIndex = ord( $c{2} ) - 0xa7;
 633                                                 if( $tIndex < 0 ) $tIndex = ord( $c{2} ) - 0x80 + (0x11c0 - 0x11a7);
 634
 635                                                 # Increment the code point by $tIndex, without
 636                                                 # the function overhead of decoding and recoding UTF-8
 637                                                 #
 638                                                 $tail = ord( $startChar{2} ) + $tIndex;
 639                                                 if( $tail > 0xbf ) {
 640                                                         $tail -= 0x40;
 641                                                         $mid = ord( $startChar{1} ) + 1;
 642                                                         if( $mid > 0xbf ) {
 643                                                                 $startChar{0} = chr( ord( $startChar{0} ) + 1 );
 644                                                                 $mid -= 0x40;
 645                                                         }
 646                                                         $startChar{1} = chr( $mid );
 647                                                 }
 648                                                 $startChar{2} = chr( $tail );
 649                                                 continue;
 650                                         }
 651                                 }
 652                         }
 653                         $out .= $startChar;
 654                         $out .= $combining;
 655                         $startChar = $c;
 656                         $combining = '';
 657                         $lastClass = 0;
 658                 }
 659                 $out .= $startChar . $combining;
 660                 return $out;
 661         }
 662
 663         /**
 664          * This is just used for the benchmark, comparing how long it takes to
 665          * interate through a string without really doing anything of substance.
 666          * @param string $string
 667          * @return string
 668          */
 669         function placebo( $string ) {
 670                 $len = strlen( $string );
 671                 $out = '';
 672                 for( $i = 0; $i < $len; $i++ ) {
 673                         $out .= $string{$i};
 674                 }
 675                 return $out;
 676         }
 677 }
 678
 679 ?>