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 # 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  18 # http://www.gnu.org/copyleft/gpl.html
  19
  20 /** */
  21 require_once dirname(__FILE__).'/UtfNormalUtil.php';
  22
  23 global $utfCombiningClass, $utfCanonicalComp, $utfCanonicalDecomp;
  24 $utfCombiningClass = NULL;
  25 $utfCanonicalComp = NULL;
  26 $utfCanonicalDecomp = NULL;
  27
  28 # Load compatibility decompositions on demand if they are needed.
  29 global $utfCompatibilityDecomp;
  30 $utfCompatibilityDecomp = NULL;
  31
  32 define( 'UNICODE_HANGUL_FIRST', 0xac00 );
  33 define( 'UNICODE_HANGUL_LAST',  0xd7a3 );
  34
  35 define( 'UNICODE_HANGUL_LBASE', 0x1100 );
  36 define( 'UNICODE_HANGUL_VBASE', 0x1161 );
  37 define( 'UNICODE_HANGUL_TBASE', 0x11a7 );
  38
  39 define( 'UNICODE_HANGUL_LCOUNT', 19 );
  40 define( 'UNICODE_HANGUL_VCOUNT', 21 );
  41 define( 'UNICODE_HANGUL_TCOUNT', 28 );
  42 define( 'UNICODE_HANGUL_NCOUNT', UNICODE_HANGUL_VCOUNT * UNICODE_HANGUL_TCOUNT );
  43
  44 define( 'UNICODE_HANGUL_LEND', UNICODE_HANGUL_LBASE + UNICODE_HANGUL_LCOUNT - 1 );
  45 define( 'UNICODE_HANGUL_VEND', UNICODE_HANGUL_VBASE + UNICODE_HANGUL_VCOUNT - 1 );
  46 define( 'UNICODE_HANGUL_TEND', UNICODE_HANGUL_TBASE + UNICODE_HANGUL_TCOUNT - 1 );
  47
  48 define( 'UNICODE_SURROGATE_FIRST', 0xd800 );
  49 define( 'UNICODE_SURROGATE_LAST', 0xdfff );
  50 define( 'UNICODE_MAX', 0x10ffff );
  51 define( 'UNICODE_REPLACEMENT', 0xfffd );
  52
  53
  54 define( 'UTF8_HANGUL_FIRST', "\xea\xb0\x80" /*codepointToUtf8( UNICODE_HANGUL_FIRST )*/ );
  55 define( 'UTF8_HANGUL_LAST', "\xed\x9e\xa3" /*codepointToUtf8( UNICODE_HANGUL_LAST )*/ );
  56
  57 define( 'UTF8_HANGUL_LBASE', "\xe1\x84\x80" /*codepointToUtf8( UNICODE_HANGUL_LBASE )*/ );
  58 define( 'UTF8_HANGUL_VBASE', "\xe1\x85\xa1" /*codepointToUtf8( UNICODE_HANGUL_VBASE )*/ );
  59 define( 'UTF8_HANGUL_TBASE', "\xe1\x86\xa7" /*codepointToUtf8( UNICODE_HANGUL_TBASE )*/ );
  60
  61 define( 'UTF8_HANGUL_LEND', "\xe1\x84\x92" /*codepointToUtf8( UNICODE_HANGUL_LEND )*/ );
  62 define( 'UTF8_HANGUL_VEND', "\xe1\x85\xb5" /*codepointToUtf8( UNICODE_HANGUL_VEND )*/ );
  63 define( 'UTF8_HANGUL_TEND', "\xe1\x87\x82" /*codepointToUtf8( UNICODE_HANGUL_TEND )*/ );
  64
  65 define( 'UTF8_SURROGATE_FIRST', "\xed\xa0\x80" /*codepointToUtf8( UNICODE_SURROGATE_FIRST )*/ );
  66 define( 'UTF8_SURROGATE_LAST', "\xed\xbf\xbf" /*codepointToUtf8( UNICODE_SURROGATE_LAST )*/ );
  67 define( 'UTF8_MAX', "\xf4\x8f\xbf\xbf" /*codepointToUtf8( UNICODE_MAX )*/ );
  68 define( 'UTF8_REPLACEMENT', "\xef\xbf\xbd" /*codepointToUtf8( UNICODE_REPLACEMENT )*/ );
  69 #define( 'UTF8_REPLACEMENT', '!' );
  70
  71 define( 'UTF8_OVERLONG_A', "\xc1\xbf" );
  72 define( 'UTF8_OVERLONG_B', "\xe0\x9f\xbf" );
  73 define( 'UTF8_OVERLONG_C', "\xf0\x8f\xbf\xbf" );
  74
  75 # These two ranges are illegal
  76 define( 'UTF8_FDD0', "\xef\xb7\x90" /*codepointToUtf8( 0xfdd0 )*/ );
  77 define( 'UTF8_FDEF', "\xef\xb7\xaf" /*codepointToUtf8( 0xfdef )*/ );
  78 define( 'UTF8_FFFE', "\xef\xbf\xbe" /*codepointToUtf8( 0xfffe )*/ );
  79 define( 'UTF8_FFFF', "\xef\xbf\xbf" /*codepointToUtf8( 0xffff )*/ );
  80
  81 define( 'UTF8_HEAD', false );
  82 define( 'UTF8_TAIL', true );
  83
  84
  85 /**
  86  * For using the ICU wrapper
  87  */
  88 define( 'UNORM_NONE', 1 );
  89 define( 'UNORM_NFD',  2 );
  90 define( 'UNORM_NFKD', 3 );
  91 define( 'UNORM_NFC',  4 );
  92 define( 'UNORM_DEFAULT', UNORM_NFC );
  93 define( 'UNORM_NFKC', 5 );
  94 define( 'UNORM_FCD',  6 );
  95
  96 define( 'NORMALIZE_ICU', function_exists( 'utf8_normalize' ) );
  97
  98 /**
  99  * Unicode normalization routines for working with UTF-8 strings.
 100  * Currently assumes that input strings are valid UTF-8!
 101  *
 102  * Not as fast as I'd like, but should be usable for most purposes.
 103  * UtfNormal::toNFC() will bail early if given ASCII text or text
 104  * it can quickly deterimine is already normalized.
 105  *
 106  * All functions can be called static.
 107  *
 108  * See description of forms at http://www.unicode.org/reports/tr15/
 109  *
 110  * @addtogroup UtfNormal
 111  */
 112 class UtfNormal {
 113         /**
 114          * The ultimate convenience function! Clean up invalid UTF-8 sequences,
 115          * and convert to normal form C, canonical composition.
 116          *
 117          * Fast return for pure ASCII strings; some lesser optimizations for
 118          * strings containing only known-good characters. Not as fast as toNFC().
 119          *
 120          * @param string $string a UTF-8 string
 121          * @return string a clean, shiny, normalized UTF-8 string
 122          * @static
 123          */
 124         static function cleanUp( $string ) {
 125                 if( NORMALIZE_ICU ) {
 126                         # We exclude a few chars that ICU would not.
 127                         $string = preg_replace(
 128                                 '/[\x00-\x08\x0b\x0c\x0e-\x1f]/',
 129                                 UTF8_REPLACEMENT,
 130                                 $string );
 131                         $string = str_replace( UTF8_FFFE, UTF8_REPLACEMENT, $string );
 132                         $string = str_replace( UTF8_FFFF, UTF8_REPLACEMENT, $string );
 133
 134                         # UnicodeString constructor fails if the string ends with a
 135                         # head byte. Add a junk char at the end, we'll strip it off.
 136                         return rtrim( utf8_normalize( $string . "\x01", UNORM_NFC ), "\x01" );
 137                 } elseif( UtfNormal::quickIsNFCVerify( $string ) ) {
 138                         # Side effect -- $string has had UTF-8 errors cleaned up.
 139                         return $string;
 140                 } else {
 141                         return UtfNormal::NFC( $string );
 142                 }
 143         }
 144
 145         /**
 146          * Convert a UTF-8 string to normal form C, canonical composition.
 147          * Fast return for pure ASCII strings; some lesser optimizations for
 148          * strings containing only known-good characters.
 149          *
 150          * @param string $string a valid UTF-8 string. Input is not validated.
 151          * @return string a UTF-8 string in normal form C
 152          * @static
 153          */
 154         static function toNFC( $string ) {
 155                 if( NORMALIZE_ICU )
 156                         return utf8_normalize( $string, UNORM_NFC );
 157                 elseif( UtfNormal::quickIsNFC( $string ) )
 158                         return $string;
 159                 else
 160                         return UtfNormal::NFC( $string );
 161         }
 162
 163         /**
 164          * Convert a UTF-8 string to normal form D, canonical decomposition.
 165          * Fast return for pure ASCII strings.
 166          *
 167          * @param string $string a valid UTF-8 string. Input is not validated.
 168          * @return string a UTF-8 string in normal form D
 169          * @static
 170          */
 171         static function toNFD( $string ) {
 172                 if( NORMALIZE_ICU )
 173                         return utf8_normalize( $string, UNORM_NFD );
 174                 elseif( preg_match( '/[\x80-\xff]/', $string ) )
 175                         return UtfNormal::NFD( $string );
 176                 else
 177                         return $string;
 178         }
 179
 180         /**
 181          * Convert a UTF-8 string to normal form KC, compatibility composition.
 182          * This may cause irreversible information loss, use judiciously.
 183          * Fast return for pure ASCII strings.
 184          *
 185          * @param string $string a valid UTF-8 string. Input is not validated.
 186          * @return string a UTF-8 string in normal form KC
 187          * @static
 188          */
 189         static function toNFKC( $string ) {
 190                 if( NORMALIZE_ICU )
 191                         return utf8_normalize( $string, UNORM_NFKC );
 192                 elseif( preg_match( '/[\x80-\xff]/', $string ) )
 193                         return UtfNormal::NFKC( $string );
 194                 else
 195                         return $string;
 196         }
 197
 198         /**
 199          * Convert a UTF-8 string to normal form KD, compatibility decomposition.
 200          * This may cause irreversible information loss, use judiciously.
 201          * Fast return for pure ASCII strings.
 202          *
 203          * @param string $string a valid UTF-8 string. Input is not validated.
 204          * @return string a UTF-8 string in normal form KD
 205          * @static
 206          */
 207         static function toNFKD( $string ) {
 208                 if( NORMALIZE_ICU )
 209                         return utf8_normalize( $string, UNORM_NFKD );
 210                 elseif( preg_match( '/[\x80-\xff]/', $string ) )
 211                         return UtfNormal::NFKD( $string );
 212                 else
 213                         return $string;
 214         }
 215
 216         /**
 217          * Load the basic composition data if necessary
 218          * @private
 219          * @static
 220          */
 221         static function loadData() {
 222                 global $utfCombiningClass;
 223                 if( !isset( $utfCombiningClass ) ) {
 224                         require_once( dirname(__FILE__) . '/UtfNormalData.inc' );
 225                 }
 226         }
 227
 228         /**
 229          * Returns true if the string is _definitely_ in NFC.
 230          * Returns false if not or uncertain.
 231          * @param string $string a valid UTF-8 string. Input is not validated.
 232          * @return bool
 233          * @static
 234          */
 235         static function quickIsNFC( $string ) {
 236                 # ASCII is always valid NFC!
 237                 # If it's pure ASCII, let it through.
 238                 if( !preg_match( '/[\x80-\xff]/', $string ) ) return true;
 239
 240                 UtfNormal::loadData();
 241                 global $utfCheckNFC, $utfCombiningClass;
 242                 $len = strlen( $string );
 243                 for( $i = 0; $i < $len; $i++ ) {
 244                         $c = $string{$i};
 245                         $n = ord( $c );
 246                         if( $n < 0x80 ) {
 247                                 continue;
 248                         } elseif( $n >= 0xf0 ) {
 249                                 $c = substr( $string, $i, 4 );
 250                                 $i += 3;
 251                         } elseif( $n >= 0xe0 ) {
 252                                 $c = substr( $string, $i, 3 );
 253                                 $i += 2;
 254                         } elseif( $n >= 0xc0 ) {
 255                                 $c = substr( $string, $i, 2 );
 256                                 $i++;
 257                         }
 258                         if( isset( $utfCheckNFC[$c] ) ) {
 259                                 # If it's NO or MAYBE, bail and do the slow check.
 260                                 return false;
 261                         }
 262                         if( isset( $utfCombiningClass[$c] ) ) {
 263                                 # Combining character? We might have to do sorting, at least.
 264                                 return false;
 265                         }
 266                 }
 267                 return true;
 268         }
 269
 270         /**
 271          * Returns true if the string is _definitely_ in NFC.
 272          * Returns false if not or uncertain.
 273          * @param string $string a UTF-8 string, altered on output to be valid UTF-8 safe for XML.
 274          * @static
 275          */
 276         static function quickIsNFCVerify( &$string ) {
 277                 # Screen out some characters that eg won't be allowed in XML
 278                 $string = preg_replace( '/[\x00-\x08\x0b\x0c\x0e-\x1f]/', UTF8_REPLACEMENT, $string );
 279
 280                 # ASCII is always valid NFC!
 281                 # If we're only ever given plain ASCII, we can avoid the overhead
 282                 # of initializing the decomposition tables by skipping out early.
 283                 if( !preg_match( '/[\x80-\xff]/', $string ) ) return true;
 284
 285                 static $checkit = null, $tailBytes = null, $utfCheckOrCombining = null;
 286                 if( !isset( $checkit ) ) {
 287                         # Load/build some scary lookup tables...
 288                         UtfNormal::loadData();
 289                         global $utfCheckNFC, $utfCombiningClass;
 290
 291                         $utfCheckOrCombining = array_merge( $utfCheckNFC, $utfCombiningClass );
 292
 293                         # Head bytes for sequences which we should do further validity checks
 294                         $checkit = array_flip( array_map( 'chr',
 295                                         array( 0xc0, 0xc1, 0xe0, 0xed, 0xef,
 296                                                    0xf0, 0xf1, 0xf2, 0xf3, 0xf4, 0xf5, 0xf6, 0xf7,
 297                                                    0xf8, 0xf9, 0xfa, 0xfb, 0xfc, 0xfd, 0xfe, 0xff ) ) );
 298
 299                         # Each UTF-8 head byte is followed by a certain
 300                         # number of tail bytes.
 301                         $tailBytes = array();
 302                         for( $n = 0; $n < 256; $n++ ) {
 303                                 if( $n < 0xc0 ) {
 304                                         $remaining = 0;
 305                                 } elseif( $n < 0xe0 ) {
 306                                         $remaining = 1;
 307                                 } elseif( $n < 0xf0 ) {
 308                                         $remaining = 2;
 309                                 } elseif( $n < 0xf8 ) {
 310                                         $remaining = 3;
 311                                 } elseif( $n < 0xfc ) {
 312                                         $remaining = 4;
 313                                 } elseif( $n < 0xfe ) {
 314                                         $remaining = 5;
 315                                 } else {
 316                                         $remaining = 0;
 317                                 }
 318                                 $tailBytes[chr($n)] = $remaining;
 319                         }
 320                 }
 321
 322                 # Chop the text into pure-ASCII and non-ASCII areas;
 323                 # large ASCII parts can be handled much more quickly.
 324                 # Don't chop up Unicode areas for punctuation, though,
 325                 # that wastes energy.
 326                 $matches = array();
 327                 preg_match_all(
 328                         '/([\x00-\x7f]+|[\x80-\xff][\x00-\x40\x5b-\x5f\x7b-\xff]*)/',
 329                         $string, $matches );
 330
 331                 $looksNormal = true;
 332                 $base = 0;
 333                 $replace = array();
 334                 foreach( $matches[1] as $str ) {
 335                         $chunk = strlen( $str );
 336
 337                         if( $str{0} < "\x80" ) {
 338                                 # ASCII chunk: guaranteed to be valid UTF-8
 339                                 # and in normal form C, so skip over it.
 340                                 $base += $chunk;
 341                                 continue;
 342                         }
 343
 344                         # We'll have to examine the chunk byte by byte to ensure
 345                         # that it consists of valid UTF-8 sequences, and to see
 346                         # if any of them might not be normalized.
 347                         #
 348                         # Since PHP is not the fastest language on earth, some of
 349                         # this code is a little ugly with inner loop optimizations.
 350
 351                         $head = '';
 352                         $len = $chunk + 1; # Counting down is faster. I'm *so* sorry.
 353
 354                         for( $i = -1; --$len; ) {
 355                                 if( $remaining = $tailBytes[$c = $str{++$i}] ) {
 356                                         # UTF-8 head byte!
 357                                         $sequence = $head = $c;
 358                                         do {
 359                                                 # Look for the defined number of tail bytes...
 360                                                 if( --$len && ( $c = $str{++$i} ) >= "\x80" && $c < "\xc0" ) {
 361                                                         # Legal tail bytes are nice.
 362                                                         $sequence .= $c;
 363                                                 } else {
 364                                                         if( 0 == $len ) {
 365                                                                 # Premature end of string!
 366                                                                 # Drop a replacement character into output to
 367                                                                 # represent the invalid UTF-8 sequence.
 368                                                                 $replace[] = array( UTF8_REPLACEMENT,
 369                                                                                                         $base + $i + 1 - strlen( $sequence ),
 370                                                                                                         strlen( $sequence ) );
 371                                                                 break 2;
 372                                                         } else {
 373                                                                 # Illegal tail byte; abandon the sequence.
 374                                                                 $replace[] = array( UTF8_REPLACEMENT,
 375                                                                                                         $base + $i - strlen( $sequence ),
 376                                                                                                         strlen( $sequence ) );
 377                                                                 # Back up and reprocess this byte; it may itself
 378                                                                 # be a legal ASCII or UTF-8 sequence head.
 379                                                                 --$i;
 380                                                                 ++$len;
 381                                                                 continue 2;
 382                                                         }
 383                                                 }
 384                                         } while( --$remaining );
 385
 386                                         if( isset( $checkit[$head] ) ) {
 387                                                 # Do some more detailed validity checks, for
 388                                                 # invalid characters and illegal sequences.
 389                                                 if( $head == "\xed" ) {
 390                                                         # 0xed is relatively frequent in Korean, which
 391                                                         # abuts the surrogate area, so we're doing
 392                                                         # this check separately to speed things up.
 393
 394                                                         if( $sequence >= UTF8_SURROGATE_FIRST ) {
 395                                                                 # Surrogates are legal only in UTF-16 code.
 396                                                                 # They are totally forbidden here in UTF-8
 397                                                                 # utopia.
 398                                                                 $replace[] = array( UTF8_REPLACEMENT,
 399                                                                              $base + $i + 1 - strlen( $sequence ),
 400                                                                              strlen( $sequence ) );
 401                                                                 $head = '';
 402                                                                 continue;
 403                                                         }
 404                                                 } else {
 405                                                         # Slower, but rarer checks...
 406                                                         $n = ord( $head );
 407                                                         if(
 408                                                                 # "Overlong sequences" are those that are syntactically
 409                                                                 # correct but use more UTF-8 bytes than are necessary to
 410                                                                 # encode a character. Naïve string comparisons can be
 411                                                                 # tricked into failing to see a match for an ASCII
 412                                                                 # character, for instance, which can be a security hole
 413                                                                 # if blacklist checks are being used.
 414                                                                ($n  < 0xc2 && $sequence <= UTF8_OVERLONG_A)
 415                                                                 || ($n == 0xe0 && $sequence <= UTF8_OVERLONG_B)
 416                                                                 || ($n == 0xf0 && $sequence <= UTF8_OVERLONG_C)
 417
 418                                                                 # U+FFFE and U+FFFF are explicitly forbidden in Unicode.
 419                                                                 || ($n == 0xef &&
 420                                                                            ($sequence == UTF8_FFFE)
 421                                                                         || ($sequence == UTF8_FFFF) )
 422
 423                                                                 # Unicode has been limited to 21 bits; longer
 424                                                                 # sequences are not allowed.
 425                                                                 || ($n >= 0xf0 && $sequence > UTF8_MAX) ) {
 426
 427                                                                 $replace[] = array( UTF8_REPLACEMENT,
 428                                                                                     $base + $i + 1 - strlen( $sequence ),
 429                                                                                     strlen( $sequence ) );
 430                                                                 $head = '';
 431                                                                 continue;
 432                                                         }
 433                                                 }
 434                                         }
 435
 436                                         if( isset( $utfCheckOrCombining[$sequence] ) ) {
 437                                                 # If it's NO or MAYBE, we'll have to rip
 438                                                 # the string apart and put it back together.
 439                                                 # That's going to be mighty slow.
 440                                                 $looksNormal = false;
 441                                         }
 442
 443                                         # The sequence is legal!
 444                                         $head = '';
 445                                 } elseif( $c < "\x80" ) {
 446                                         # ASCII byte.
 447                                         $head = '';
 448                                 } elseif( $c < "\xc0" ) {
 449                                         # Illegal tail bytes
 450                                         if( $head == '' ) {
 451                                                 # Out of the blue!
 452                                                 $replace[] = array( UTF8_REPLACEMENT, $base + $i, 1 );
 453                                         } else {
 454                                                 # Don't add if we're continuing a broken sequence;
 455                                                 # we already put a replacement character when we looked
 456                                                 # at the broken sequence.
 457                                                 $replace[] = array( '', $base + $i, 1 );
 458                                         }
 459                                 } else {
 460                                         # Miscellaneous freaks.
 461                                         $replace[] = array( UTF8_REPLACEMENT, $base + $i, 1 );
 462                                         $head = '';
 463                                 }
 464                         }
 465                         $base += $chunk;
 466                 }
 467                 if( count( $replace ) ) {
 468                         # There were illegal UTF-8 sequences we need to fix up.
 469                         $out = '';
 470                         $last = 0;
 471                         foreach( $replace as $rep ) {
 472                                 list( $replacement, $start, $length ) = $rep;
 473                                 if( $last < $start ) {
 474                                         $out .= substr( $string, $last, $start - $last );
 475                                 }
 476                                 $out .= $replacement;
 477                                 $last = $start + $length;
 478                         }
 479                         if( $last < strlen( $string ) ) {
 480                                 $out .= substr( $string, $last );
 481                         }
 482                         $string = $out;
 483                 }
 484                 return $looksNormal;
 485         }
 486
 487         # These take a string and run the normalization on them, without
 488         # checking for validity or any optimization etc. Input must be
 489         # VALID UTF-8!
 490         /**
 491          * @param string $string
 492          * @return string
 493          * @private
 494          * @static
 495          */
 496         static function NFC( $string ) {
 497                 return UtfNormal::fastCompose( UtfNormal::NFD( $string ) );
 498         }
 499
 500         /**
 501          * @param string $string
 502          * @return string
 503          * @private
 504          * @static
 505          */
 506         static function NFD( $string ) {
 507                 UtfNormal::loadData();
 508                 global $utfCanonicalDecomp;
 509                 return UtfNormal::fastCombiningSort(
 510                         UtfNormal::fastDecompose( $string, $utfCanonicalDecomp ) );
 511         }
 512
 513         /**
 514          * @param string $string
 515          * @return string
 516          * @private
 517          * @static
 518          */
 519         static function NFKC( $string ) {
 520                 return UtfNormal::fastCompose( UtfNormal::NFKD( $string ) );
 521         }
 522
 523         /**
 524          * @param string $string
 525          * @return string
 526          * @private
 527          * @static
 528          */
 529         static function NFKD( $string ) {
 530                 global $utfCompatibilityDecomp;
 531                 if( !isset( $utfCompatibilityDecomp ) ) {
 532                         require_once( 'UtfNormalDataK.inc' );
 533                 }
 534                 return UtfNormal::fastCombiningSort(
 535                         UtfNormal::fastDecompose( $string, $utfCompatibilityDecomp ) );
 536         }
 537
 538
 539         /**
 540          * Perform decomposition of a UTF-8 string into either D or KD form
 541          * (depending on which decomposition map is passed to us).
 542          * Input is assumed to be *valid* UTF-8. Invalid code will break.
 543          * @private
 544          * @param string $string Valid UTF-8 string
 545          * @param array $map hash of expanded decomposition map
 546          * @return string a UTF-8 string decomposed, not yet normalized (needs sorting)
 547          * @static
 548          */
 549         static function fastDecompose( $string, $map ) {
 550                 UtfNormal::loadData();
 551                 $len = strlen( $string );
 552                 $out = '';
 553                 for( $i = 0; $i < $len; $i++ ) {
 554                         $c = $string{$i};
 555                         $n = ord( $c );
 556                         if( $n < 0x80 ) {
 557                                 # ASCII chars never decompose
 558                                 # THEY ARE IMMORTAL
 559                                 $out .= $c;
 560                                 continue;
 561                         } elseif( $n >= 0xf0 ) {
 562                                 $c = substr( $string, $i, 4 );
 563                                 $i += 3;
 564                         } elseif( $n >= 0xe0 ) {
 565                                 $c = substr( $string, $i, 3 );
 566                                 $i += 2;
 567                         } elseif( $n >= 0xc0 ) {
 568                                 $c = substr( $string, $i, 2 );
 569                                 $i++;
 570                         }
 571                         if( isset( $map[$c] ) ) {
 572                                 $out .= $map[$c];
 573                                 continue;
 574                         } else {
 575                                 if( $c >= UTF8_HANGUL_FIRST && $c <= UTF8_HANGUL_LAST ) {
 576                                         # Decompose a hangul syllable into jamo;
 577                                         # hardcoded for three-byte UTF-8 sequence.
 578                                         # A lookup table would be slightly faster,
 579                                         # but adds a lot of memory & disk needs.
 580                                         #
 581                                         $index = ( (ord( $c{0} ) & 0x0f) << 12
 582                                                  | (ord( $c{1} ) & 0x3f) <<  6
 583                                                  | (ord( $c{2} ) & 0x3f) )
 584                                                - UNICODE_HANGUL_FIRST;
 585                                         $l = intval( $index / UNICODE_HANGUL_NCOUNT );
 586                                         $v = intval( ($index % UNICODE_HANGUL_NCOUNT) / UNICODE_HANGUL_TCOUNT);
 587                                         $t = $index % UNICODE_HANGUL_TCOUNT;
 588                                         $out .= "\xe1\x84" . chr( 0x80 + $l ) . "\xe1\x85" . chr( 0xa1 + $v );
 589                                         if( $t >= 25 ) {
 590                                                 $out .= "\xe1\x87" . chr( 0x80 + $t - 25 );
 591                                         } elseif( $t ) {
 592                                                 $out .= "\xe1\x86" . chr( 0xa7 + $t );
 593                                         }
 594                                         continue;
 595                                 }
 596                         }
 597                         $out .= $c;
 598                 }
 599                 return $out;
 600         }
 601
 602         /**
 603          * Sorts combining characters into canonical order. This is the
 604          * final step in creating decomposed normal forms D and KD.
 605          * @private
 606          * @param string $string a valid, decomposed UTF-8 string. Input is not validated.
 607          * @return string a UTF-8 string with combining characters sorted in canonical order
 608          * @static
 609          */
 610         static function fastCombiningSort( $string ) {
 611                 UtfNormal::loadData();
 612                 global $utfCombiningClass;
 613                 $len = strlen( $string );
 614                 $out = '';
 615                 $combiners = array();
 616                 $lastClass = -1;
 617                 for( $i = 0; $i < $len; $i++ ) {
 618                         $c = $string{$i};
 619                         $n = ord( $c );
 620                         if( $n >= 0x80 ) {
 621                                 if( $n >= 0xf0 ) {
 622                                         $c = substr( $string, $i, 4 );
 623                                         $i += 3;
 624                                 } elseif( $n >= 0xe0 ) {
 625                                         $c = substr( $string, $i, 3 );
 626                                         $i += 2;
 627                                 } elseif( $n >= 0xc0 ) {
 628                                         $c = substr( $string, $i, 2 );
 629                                         $i++;
 630                                 }
 631                                 if( isset( $utfCombiningClass[$c] ) ) {
 632                                         $lastClass = $utfCombiningClass[$c];
 633                                         if( isset( $combiners[$lastClass] ) ) {
 634                                                 $combiners[$lastClass] .= $c;
 635                                         } else {
 636                                                 $combiners[$lastClass] = $c;
 637                                         }
 638                                         continue;
 639                                 }
 640                         }
 641                         if( $lastClass ) {
 642                                 ksort( $combiners );
 643                                 $out .= implode( '', $combiners );
 644                                 $combiners = array();
 645                         }
 646                         $out .= $c;
 647                         $lastClass = 0;
 648                 }
 649                 if( $lastClass ) {
 650                         ksort( $combiners );
 651                         $out .= implode( '', $combiners );
 652                 }
 653                 return $out;
 654         }
 655
 656         /**
 657          * Produces canonically composed sequences, i.e. normal form C or KC.
 658          *
 659          * @private
 660          * @param string $string a valid UTF-8 string in sorted normal form D or KD. Input is not validated.
 661          * @return string a UTF-8 string with canonical precomposed characters used where possible
 662          * @static
 663          */
 664         static function fastCompose( $string ) {
 665                 UtfNormal::loadData();
 666                 global $utfCanonicalComp, $utfCombiningClass;
 667                 $len = strlen( $string );
 668                 $out = '';
 669                 $lastClass = -1;
 670                 $lastHangul = 0;
 671                 $startChar = '';
 672                 $combining = '';
 673                 $x1 = ord(substr(UTF8_HANGUL_VBASE,0,1));
 674                 $x2 = ord(substr(UTF8_HANGUL_TEND,0,1));
 675                 for( $i = 0; $i < $len; $i++ ) {
 676                         $c = $string{$i};
 677                         $n = ord( $c );
 678                         if( $n < 0x80 ) {
 679                                 # No combining characters here...
 680                                 $out .= $startChar;
 681                                 $out .= $combining;
 682                                 $startChar = $c;
 683                                 $combining = '';
 684                                 $lastClass = 0;
 685                                 continue;
 686                         } elseif( $n >= 0xf0 ) {
 687                                 $c = substr( $string, $i, 4 );
 688                                 $i += 3;
 689                         } elseif( $n >= 0xe0 ) {
 690                                 $c = substr( $string, $i, 3 );
 691                                 $i += 2;
 692                         } elseif( $n >= 0xc0 ) {
 693                                 $c = substr( $string, $i, 2 );
 694                                 $i++;
 695                         }
 696                         $pair = $startChar . $c;
 697                         if( $n > 0x80 ) {
 698                                 if( isset( $utfCombiningClass[$c] ) ) {
 699                                         # A combining char; see what we can do with it
 700                                         $class = $utfCombiningClass[$c];
 701                                         if( !empty( $startChar ) &&
 702                                                 $lastClass < $class &&
 703                                                 $class > 0 &&
 704                                                 isset( $utfCanonicalComp[$pair] ) ) {
 705                                                 $startChar = $utfCanonicalComp[$pair];
 706                                                 $class = 0;
 707                                         } else {
 708                                                 $combining .= $c;
 709                                         }
 710                                         $lastClass = $class;
 711                                         $lastHangul = 0;
 712                                         continue;
 713                                 }
 714                         }
 715                         # New start char
 716                         if( $lastClass == 0 ) {
 717                                 if( isset( $utfCanonicalComp[$pair] ) ) {
 718                                         $startChar = $utfCanonicalComp[$pair];
 719                                         $lastHangul = 0;
 720                                         continue;
 721                                 }
 722                                 if( $n >= $x1 && $n <= $x2 ) {
 723                                         # WARNING: Hangul code is painfully slow.
 724                                         # I apologize for this ugly, ugly code; however
 725                                         # performance is even more teh suck if we call
 726                                         # out to nice clean functions. Lookup tables are
 727                                         # marginally faster, but require a lot of space.
 728                                         #
 729                                         if( $c >= UTF8_HANGUL_VBASE &&
 730                                                 $c <= UTF8_HANGUL_VEND &&
 731                                                 $startChar >= UTF8_HANGUL_LBASE &&
 732                                                 $startChar <= UTF8_HANGUL_LEND ) {
 733                                                 #
 734                                                 #$lIndex = utf8ToCodepoint( $startChar ) - UNICODE_HANGUL_LBASE;
 735                                                 #$vIndex = utf8ToCodepoint( $c ) - UNICODE_HANGUL_VBASE;
 736                                                 $lIndex = ord( $startChar{2} ) - 0x80;
 737                                                 $vIndex = ord( $c{2}         ) - 0xa1;
 738
 739                                                 $hangulPoint = UNICODE_HANGUL_FIRST +
 740                                                         UNICODE_HANGUL_TCOUNT *
 741                                                         (UNICODE_HANGUL_VCOUNT * $lIndex + $vIndex);
 742
 743                                                 # Hardcode the limited-range UTF-8 conversion:
 744                                                 $startChar = chr( $hangulPoint >> 12 & 0x0f | 0xe0 ) .
 745                                                                          chr( $hangulPoint >>  6 & 0x3f | 0x80 ) .
 746                                                                          chr( $hangulPoint       & 0x3f | 0x80 );
 747                                                 $lastHangul = 0;
 748                                                 continue;
 749                                         } elseif( $c >= UTF8_HANGUL_TBASE &&
 750                                                           $c <= UTF8_HANGUL_TEND &&
 751                                                           $startChar >= UTF8_HANGUL_FIRST &&
 752                                                           $startChar <= UTF8_HANGUL_LAST &&
 753                                                           !$lastHangul ) {
 754                                                 # $tIndex = utf8ToCodepoint( $c ) - UNICODE_HANGUL_TBASE;
 755                                                 $tIndex = ord( $c{2} ) - 0xa7;
 756                                                 if( $tIndex < 0 ) $tIndex = ord( $c{2} ) - 0x80 + (0x11c0 - 0x11a7);
 757
 758                                                 # Increment the code point by $tIndex, without
 759                                                 # the function overhead of decoding and recoding UTF-8
 760                                                 #
 761                                                 $tail = ord( $startChar{2} ) + $tIndex;
 762                                                 if( $tail > 0xbf ) {
 763                                                         $tail -= 0x40;
 764                                                         $mid = ord( $startChar{1} ) + 1;
 765                                                         if( $mid > 0xbf ) {
 766                                                                 $startChar{0} = chr( ord( $startChar{0} ) + 1 );
 767                                                                 $mid -= 0x40;
 768                                                         }
 769                                                         $startChar{1} = chr( $mid );
 770                                                 }
 771                                                 $startChar{2} = chr( $tail );
 772
 773                                                 # If there's another jamo char after this, *don't* try to merge it.
 774                                                 $lastHangul = 1;
 775                                                 continue;
 776                                         }
 777                                 }
 778                         }
 779                         $out .= $startChar;
 780                         $out .= $combining;
 781                         $startChar = $c;
 782                         $combining = '';
 783                         $lastClass = 0;
 784                         $lastHangul = 0;
 785                 }
 786                 $out .= $startChar . $combining;
 787                 return $out;
 788         }
 789
 790         /**
 791          * This is just used for the benchmark, comparing how long it takes to
 792          * interate through a string without really doing anything of substance.
 793          * @param string $string
 794          * @return string
 795          * @static
 796          */
 797         static function placebo( $string ) {
 798                 $len = strlen( $string );
 799                 $out = '';
 800                 for( $i = 0; $i < $len; $i++ ) {
 801                         $out .= $string{$i};
 802                 }
 803                 return $out;
 804         }
 805 }
 806
 807 ?>