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                 static $checkit = null, $tailBytes = null, $utfCheckOrCombining = null;
 269                 if( !isset( $checkit ) ) {
 270                         # Load/build some scary lookup tables...
 271                         UtfNormal::loadData();
 272                         global $utfCheckNFC, $utfCombiningClass;
 273
 274                         $utfCheckOrCombining = array_merge( $utfCheckNFC, $utfCombiningClass );
 275
 276                         # Head bytes for sequences which we should do further validity checks
 277                         $checkit = array_flip( array_map( 'chr',
 278                                         array( 0xc0, 0xc1, 0xe0, 0xed, 0xef,
 279                                                    0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7,
 280                                                    0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff ) ) );
 281
 282                         # Each UTF-8 head byte is followed by a certain
 283                         # number of tail bytes.
 284                         $tailBytes = array();
 285                         for( $n = 0; $n < 256; $n++ ) {
 286                                 if( $n < 0xc0 ) {
 287                                         $remaining = 0;
 288                                 } elseif( $n < 0xe0 ) {
 289                                         $remaining = 1;
 290                                 } elseif( $n < 0xf0 ) {
 291                                         $remaining = 2;
 292                                 } elseif( $n < 0xf8 ) {
 293                                         $remaining = 3;
 294                                 } elseif( $n < 0xfc ) {
 295                                         $remaining = 4;
 296                                 } elseif( $n < 0xfe ) {
 297                                         $remaining = 5;
 298                                 } else {
 299                                         $remaining = 0;
 300                                 }
 301                                 $tailBytes[chr($n)] = $remaining;
 302                         }
 303                 }
 304
 305                 # Chop the text into pure-ASCII and non-ASCII areas;
 306                 # large ASCII parts can be handled much more quickly.
 307                 # Don't chop up Unicode areas for punctuation, though,
 308                 # that wastes energy.
 309                 preg_match_all(
 310                         '/([\x00-\x7f]+|[\x80-\xff][\x00-\x40\x5b-\x5f\x7b-\xff]*)/',
 311                         $string, $matches );
 312
 313                 $looksNormal = true;
 314                 $base = 0;
 315                 $replace = array();
 316                 foreach( $matches[1] as $str ) {
 317                         $chunk = strlen( $str );
 318
 319                         if( $str{0} < "\x80" ) {
 320                                 # ASCII chunk: guaranteed to be valid UTF-8
 321                                 # and in normal form C, so skip over it.
 322                                 $base += $chunk;
 323                                 continue;
 324                         }
 325
 326                         # We'll have to examine the chunk byte by byte to ensure
 327                         # that it consists of valid UTF-8 sequences, and to see
 328                         # if any of them might not be normalized.
 329                         #
 330                         # Since PHP is not the fastest language on earth, some of
 331                         # this code is a little ugly with inner loop optimizations.
 332
 333                         $head = '';
 334                         $len = $chunk + 1; # Counting down is faster. I'm *so* sorry.
 335
 336                         for( $i = -1; --$len; ) {
 337                                 if( $remaining = $tailBytes[$c = $str{++$i}] ) {
 338                                         # UTF-8 head byte!
 339                                         $sequence = $head = $c;
 340                                         do {
 341                                                 # Look for the defined number of tail bytes...
 342                                                 if( --$len && ( $c = $str{++$i} ) >= "\x80" && $c < "\xc0" ) {
 343                                                         # Legal tail bytes are nice.
 344                                                         $sequence .= $c;
 345                                                 } else {
 346                                                         if( 0 == $len ) {
 347                                                                 # Premature end of string!
 348                                                                 # Drop a replacement character into output to
 349                                                                 # represent the invalid UTF-8 sequence.
 350                                                                 $replace[] = array( UTF8_REPLACEMENT,
 351                                                                                                         $base + $i + 1 - strlen( $sequence ),
 352                                                                                                         strlen( $sequence ) );
 353                                                                 break 2;
 354                                                         } else {
 355                                                                 # Illegal tail byte; abandon the sequence.
 356                                                                 $replace[] = array( UTF8_REPLACEMENT,
 357                                                                                                         $base + $i - strlen( $sequence ),
 358                                                                                                         strlen( $sequence ) );
 359                                                                 # Back up and reprocess this byte; it may itself
 360                                                                 # be a legal ASCII or UTF-8 sequence head.
 361                                                                 --$i;
 362                                                                 ++$len;
 363                                                                 continue 2;
 364                                                         }
 365                                                 }
 366                                         } while( --$remaining );
 367
 368                                         if( isset( $checkit[$head] ) ) {
 369                                                 # Do some more detailed validity checks, for
 370                                                 # invalid characters and illegal sequences.
 371                                                 if( $head == "\xed" ) {
 372                                                         # 0xed is relatively frequent in Korean, which
 373                                                         # abuts the surrogate area, so we're doing
 374                                                         # this check separately to speed things up.
 375
 376                                                         if( $sequence >= UTF8_SURROGATE_FIRST ) {
 377                                                                 # Surrogates are legal only in UTF-16 code.
 378                                                                 # They are totally forbidden here in UTF-8
 379                                                                 # utopia.
 380                                                                 $replace[] = array( UTF8_REPLACEMENT,
 381                                                                              $base + $i + 1 - strlen( $sequence ),
 382                                                                              strlen( $sequence ) );
 383                                                                 continue;
 384                                                         }
 385                                                 } else {
 386                                                         # Slower, but rarer checks...
 387                                                         $n = ord( $head );
 388                                                         if(
 389                                                                 # "Overlong sequences" are those that are syntactically
 390                                                                 # correct but use more UTF-8 bytes than are necessary to
 391                                                                 # encode a character. Naïve string comparisons can be
 392                                                                 # tricked into failing to see a match for an ASCII
 393                                                                 # character, for instance, which can be a security hole
 394                                                                 # if blacklist checks are being used.
 395                                                                ($n  < 0xc2 && $sequence <= UTF8_OVERLONG_A)
 396                                                                 || ($n == 0xe0 && $sequence <= UTF8_OVERLONG_B)
 397                                                                 || ($n == 0xf0 && $sequence <= UTF8_OVERLONG_C)
 398
 399                                                                 # U+FFFE and U+FFFF are explicitly forbidden in Unicode.
 400                                                                 || ($n == 0xef &&
 401                                                                            ($sequence == UTF8_FFFE)
 402                                                                         || ($sequence == UTF8_FFFF) )
 403
 404                                                                 # Unicode has been limited to 21 bits; longer
 405                                                                 # sequences are not allowed.
 406                                                                 || ($n >= 0xf0 && $sequence > UTF8_MAX) ) {
 407
 408                                                                 $replace[] = array( UTF8_REPLACEMENT,
 409                                                                                     $base + $i + 1 - strlen( $sequence ),
 410                                                                                     strlen( $sequence ) );
 411                                                                 $head = '';
 412                                                                 continue;
 413                                                         }
 414                                                 }
 415                                         }
 416
 417                                         if( isset( $utfCheckOrCombining[$sequence] ) ) {
 418                                                 # If it's NO or MAYBE, we'll have to rip
 419                                                 # the string apart and put it back together.
 420                                                 # That's going to be mighty slow.
 421                                                 $looksNormal = false;
 422                                         }
 423
 424                                         # The sequence is legal!
 425                                         $head = '';
 426                                 } elseif( $c < "\x80" ) {
 427                                         # ASCII byte.
 428                                         $head = '';
 429                                 } elseif( $c < "\xc0" ) {
 430                                         # Illegal tail bytes
 431                                         if( $head == '' ) {
 432                                                 # Out of the blue!
 433                                                 $replace[] = array( UTF8_REPLACEMENT, $base + $i, 1 );
 434                                         } else {
 435                                                 # Don't add if we're continuing a broken sequence;
 436                                                 # we already put a replacement character when we looked
 437                                                 # at the broken sequence.
 438                                                 $replace[] = array( '', $base + $i, 1 );
 439                                         }
 440                                 } else {
 441                                         # Miscellaneous freaks.
 442                                         $replace[] = array( UTF8_REPLACEMENT, $base + $i, 1 );
 443                                         $head = '';
 444                                 }
 445                         }
 446                         $base += $chunk;
 447                 }
 448                 if( count( $replace ) ) {
 449                         # There were illegal UTF-8 sequences we need to fix up.
 450                         $out = '';
 451                         $last = 0;
 452                         foreach( $replace as $rep ) {
 453                                 list( $replacement, $start, $length ) = $rep;
 454                                 if( $last < $start ) {
 455                                         $out .= substr( $string, $last, $start - $last );
 456                                 }
 457                                 $out .= $replacement;
 458                                 $last = $start + $length;
 459                         }
 460                         if( $last < strlen( $string ) ) {
 461                                 $out .= substr( $string, $last );
 462                         }
 463                         $string = $out;
 464                 }
 465                 return $looksNormal;
 466         }
 467
 468         # These take a string and run the normalization on them, without
 469         # checking for validity or any optimization etc. Input must be
 470         # VALID UTF-8!
 471         /**
 472          * @param string $string
 473          * @return string
 474          * @access private
 475          */
 476         function NFC( $string ) {
 477                 return UtfNormal::fastCompose( UtfNormal::NFD( $string ) );
 478         }
 479
 480         /**
 481          * @param string $string
 482          * @return string
 483          * @access private
 484          */
 485         function NFD( $string ) {
 486                 UtfNormal::loadData();
 487                 global $utfCanonicalDecomp;
 488                 return UtfNormal::fastCombiningSort(
 489                         UtfNormal::fastDecompose( $string, $utfCanonicalDecomp ) );
 490         }
 491
 492         /**
 493          * @param string $string
 494          * @return string
 495          * @access private
 496          */
 497         function NFKC( $string ) {
 498                 return UtfNormal::fastCompose( UtfNormal::NFKD( $string ) );
 499         }
 500
 501         /**
 502          * @param string $string
 503          * @return string
 504          * @access private
 505          */
 506         function NFKD( $string ) {
 507                 global $utfCompatibilityDecomp;
 508                 if( !isset( $utfCompatibilityDecomp ) ) {
 509                         require_once( 'UtfNormalDataK.inc' );
 510                 }
 511                 return UtfNormal::fastCombiningSort(
 512                         UtfNormal::fastDecompose( $string, $utfCompatibilityDecomp ) );
 513         }
 514
 515
 516         /**
 517          * Perform decomposition of a UTF-8 string into either D or KD form
 518          * (depending on which decomposition map is passed to us).
 519          * Input is assumed to be *valid* UTF-8. Invalid code will break.
 520          * @access private
 521          * @param string $string Valid UTF-8 string
 522          * @param array $map hash of expanded decomposition map
 523          * @return string a UTF-8 string decomposed, not yet normalized (needs sorting)
 524          */
 525         function fastDecompose( $string, &$map ) {
 526                 UtfNormal::loadData();
 527                 $len = strlen( $string );
 528                 $out = '';
 529                 for( $i = 0; $i < $len; $i++ ) {
 530                         $c = $string{$i};
 531                         $n = ord( $c );
 532                         if( $n < 0x80 ) {
 533                                 # ASCII chars never decompose
 534                                 # THEY ARE IMMORTAL
 535                                 $out .= $c;
 536                                 continue;
 537                         } elseif( $n >= 0xf0 ) {
 538                                 $c = substr( $string, $i, 4 );
 539                                 $i += 3;
 540                         } elseif( $n >= 0xe0 ) {
 541                                 $c = substr( $string, $i, 3 );
 542                                 $i += 2;
 543                         } elseif( $n >= 0xc0 ) {
 544                                 $c = substr( $string, $i, 2 );
 545                                 $i++;
 546                         }
 547                         if( isset( $map[$c] ) ) {
 548                                 $out .= $map[$c];
 549                                 continue;
 550                         } else {
 551                                 if( $c >= UTF8_HANGUL_FIRST && $c <= UTF8_HANGUL_LAST ) {
 552                                         # Decompose a hangul syllable into jamo;
 553                                         # hardcoded for three-byte UTF-8 sequence.
 554                                         # A lookup table would be slightly faster,
 555                                         # but adds a lot of memory & disk needs.
 556                                         #
 557                                         $index = ( (ord( $c{0} ) & 0x0f) << 12
 558                                                  | (ord( $c{1} ) & 0x3f) <<  6
 559                                                  | (ord( $c{2} ) & 0x3f) )
 560                                                - UNICODE_HANGUL_FIRST;
 561                                         $l = IntVal( $index / UNICODE_HANGUL_NCOUNT );
 562                                         $v = IntVal( ($index % UNICODE_HANGUL_NCOUNT) / UNICODE_HANGUL_TCOUNT);
 563                                         $t = $index % UNICODE_HANGUL_TCOUNT;
 564                                         $out .= "\xe1\x84" . chr( 0x80 + $l ) . "\xe1\x85" . chr( 0xa1 + $v );
 565                                         if( $t >= 25 ) {
 566                                                 $out .= "\xe1\x87" . chr( 0x80 + $t - 25 );
 567                                         } elseif( $t ) {
 568                                                 $out .= "\xe1\x86" . chr( 0xa7 + $t );
 569                                         }
 570                                         continue;
 571                                 }
 572                         }
 573                         $out .= $c;
 574                 }
 575                 return $out;
 576         }
 577
 578         /**
 579          * Sorts combining characters into canonical order. This is the
 580          * final step in creating decomposed normal forms D and KD.
 581          * @access private
 582          * @param string $string a valid, decomposed UTF-8 string. Input is not validated.
 583          * @return string a UTF-8 string with combining characters sorted in canonical order
 584          */
 585         function fastCombiningSort( $string ) {
 586                 UtfNormal::loadData();
 587                 global $utfCombiningClass;
 588                 $len = strlen( $string );
 589                 $out = '';
 590                 $combiners = array();
 591                 $lastClass = -1;
 592                 for( $i = 0; $i < $len; $i++ ) {
 593                         $c = $string{$i};
 594                         $n = ord( $c );
 595                         if( $n >= 0x80 ) {
 596                                 if( $n >= 0xf0 ) {
 597                                         $c = substr( $string, $i, 4 );
 598                                         $i += 3;
 599                                 } elseif( $n >= 0xe0 ) {
 600                                         $c = substr( $string, $i, 3 );
 601                                         $i += 2;
 602                                 } elseif( $n >= 0xc0 ) {
 603                                         $c = substr( $string, $i, 2 );
 604                                         $i++;
 605                                 }
 606                                 if( isset( $utfCombiningClass[$c] ) ) {
 607                                         $lastClass = $utfCombiningClass[$c];
 608                                         @$combiners[$lastClass] .= $c;
 609                                         continue;
 610                                 }
 611                         }
 612                         if( $lastClass ) {
 613                                 ksort( $combiners );
 614                                 $out .= implode( '', $combiners );
 615                                 $combiners = array();
 616                         }
 617                         $out .= $c;
 618                         $lastClass = 0;
 619                 }
 620                 if( $lastClass ) {
 621                         ksort( $combiners );
 622                         $out .= implode( '', $combiners );
 623                 }
 624                 return $out;
 625         }
 626
 627         /**
 628          * Produces canonically composed sequences, i.e. normal form C or KC.
 629          *
 630          * @access private
 631          * @param string $string a valid UTF-8 string in sorted normal form D or KD. Input is not validated.
 632          * @return string a UTF-8 string with canonical precomposed characters used where possible
 633          */
 634         function fastCompose( $string ) {
 635                 UtfNormal::loadData();
 636                 global $utfCanonicalComp, $utfCombiningClass;
 637                 $len = strlen( $string );
 638                 $out = '';
 639                 $lastClass = -1;
 640                 $startChar = '';
 641                 $combining = '';
 642                 $x1 = ord(substr(UTF8_HANGUL_VBASE,0,1));
 643                 $x2 = ord(substr(UTF8_HANGUL_TEND,0,1));
 644                 for( $i = 0; $i < $len; $i++ ) {
 645                         $c = $string{$i};
 646                         $n = ord( $c );
 647                         if( $n < 0x80 ) {
 648                                 # No combining characters here...
 649                                 $out .= $startChar;
 650                                 $out .= $combining;
 651                                 $startChar = $c;
 652                                 $combining = '';
 653                                 $lastClass = 0;
 654                                 continue;
 655                         } elseif( $n >= 0xf0 ) {
 656                                 $c = substr( $string, $i, 4 );
 657                                 $i += 3;
 658                         } elseif( $n >= 0xe0 ) {
 659                                 $c = substr( $string, $i, 3 );
 660                                 $i += 2;
 661                         } elseif( $n >= 0xc0 ) {
 662                                 $c = substr( $string, $i, 2 );
 663                                 $i++;
 664                         }
 665                         $pair = $startChar . $c;
 666                         if( $n > 0x80 ) {
 667                                 if( isset( $utfCombiningClass[$c] ) ) {
 668                                         # A combining char; see what we can do with it
 669                                         $class = $utfCombiningClass[$c];
 670                                         if( !empty( $startChar ) &&
 671                                                 $lastClass < $class &&
 672                                                 $class > 0 &&
 673                                                 isset( $utfCanonicalComp[$pair] ) ) {
 674                                                 $startChar = $utfCanonicalComp[$pair];
 675                                                 $class = 0;
 676                                         } else {
 677                                                 $combining .= $c;
 678                                         }
 679                                         $lastClass = $class;
 680                                         continue;
 681                                 }
 682                         }
 683                         # New start char
 684                         if( $lastClass == 0 ) {
 685                                 if( isset( $utfCanonicalComp[$pair] ) ) {
 686                                         $startChar = $utfCanonicalComp[$pair];
 687                                         continue;
 688                                 }
 689                                 if( $n >= $x1 && $n <= $x2 ) {
 690                                         # WARNING: Hangul code is painfully slow.
 691                                         # I apologize for this ugly, ugly code; however
 692                                         # performance is even more teh suck if we call
 693                                         # out to nice clean functions. Lookup tables are
 694                                         # marginally faster, but require a lot of space.
 695                                         #
 696                                         if( $c >= UTF8_HANGUL_VBASE &&
 697                                                 $c <= UTF8_HANGUL_VEND &&
 698                                                 $startChar >= UTF8_HANGUL_LBASE &&
 699                                                 $startChar <= UTF8_HANGUL_LEND ) {
 700                                                 #
 701                                                 #$lIndex = utf8ToCodepoint( $startChar ) - UNICODE_HANGUL_LBASE;
 702                                                 #$vIndex = utf8ToCodepoint( $c ) - UNICODE_HANGUL_VBASE;
 703                                                 $lIndex = ord( $startChar{2} ) - 0x80;
 704                                                 $vIndex = ord( $c{2}         ) - 0xa1;
 705
 706                                                 $hangulPoint = UNICODE_HANGUL_FIRST +
 707                                                         UNICODE_HANGUL_TCOUNT *
 708                                                         (UNICODE_HANGUL_VCOUNT * $lIndex + $vIndex);
 709
 710                                                 # Hardcode the limited-range UTF-8 conversion:
 711                                                 $startChar = chr( $hangulPoint >> 12 & 0x0f | 0xe0 ) .
 712                                                                          chr( $hangulPoint >>  6 & 0x3f | 0x80 ) .
 713                                                                          chr( $hangulPoint       & 0x3f | 0x80 );
 714                                                 continue;
 715                                         } elseif( $c >= UTF8_HANGUL_TBASE &&
 716                                                           $c <= UTF8_HANGUL_TEND &&
 717                                                           $startChar >= UTF8_HANGUL_FIRST &&
 718                                                           $startChar <= UTF8_HANGUL_LAST ) {
 719                                                 # $tIndex = utf8ToCodepoint( $c ) - UNICODE_HANGUL_TBASE;
 720                                                 $tIndex = ord( $c{2} ) - 0xa7;
 721                                                 if( $tIndex < 0 ) $tIndex = ord( $c{2} ) - 0x80 + (0x11c0 - 0x11a7);
 722
 723                                                 # Increment the code point by $tIndex, without
 724                                                 # the function overhead of decoding and recoding UTF-8
 725                                                 #
 726                                                 $tail = ord( $startChar{2} ) + $tIndex;
 727                                                 if( $tail > 0xbf ) {
 728                                                         $tail -= 0x40;
 729                                                         $mid = ord( $startChar{1} ) + 1;
 730                                                         if( $mid > 0xbf ) {
 731                                                                 $startChar{0} = chr( ord( $startChar{0} ) + 1 );
 732                                                                 $mid -= 0x40;
 733                                                         }
 734                                                         $startChar{1} = chr( $mid );
 735                                                 }
 736                                                 $startChar{2} = chr( $tail );
 737                                                 continue;
 738                                         }
 739                                 }
 740                         }
 741                         $out .= $startChar;
 742                         $out .= $combining;
 743                         $startChar = $c;
 744                         $combining = '';
 745                         $lastClass = 0;
 746                 }
 747                 $out .= $startChar . $combining;
 748                 return $out;
 749         }
 750
 751         /**
 752          * This is just used for the benchmark, comparing how long it takes to
 753          * interate through a string without really doing anything of substance.
 754          * @param string $string
 755          * @return string
 756          */
 757         function placebo( $string ) {
 758                 $len = strlen( $string );
 759                 $out = '';
 760                 for( $i = 0; $i < $len; $i++ ) {
 761                         $out .= $string{$i};
 762                 }
 763                 return $out;
 764         }
 765 }
 766
 767 ?>