includes/collation/IcuCollation.php

   1 <?php
   2 /**
   3  * This program is free software; you can redistribute it and/or modify
   4  * it under the terms of the GNU General Public License as published by
   5  * the Free Software Foundation; either version 2 of the License, or
   6  * (at your option) any later version.
   7  *
   8  * This program is distributed in the hope that it will be useful,
   9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  11  * GNU General Public License for more details.
  12  *
  13  * You should have received a copy of the GNU General Public License along
  14  * with this program; if not, write to the Free Software Foundation, Inc.,
  15  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  16  * http://www.gnu.org/copyleft/gpl.html
  17  *
  18  * @file
  19  */
  20
  21 /**
  22  * @since 1.16.3
  23  */
  24 class IcuCollation extends Collation {
  25         const FIRST_LETTER_VERSION = 2;
  26
  27         /** @var Collator */
  28         private $primaryCollator;
  29
  30         /** @var Collator */
  31         private $mainCollator;
  32
  33         /** @var string */
  34         private $locale;
  35
  36         /** @var Language */
  37         protected $digitTransformLanguage;
  38
  39         /** @var array */
  40         private $firstLetterData;
  41
  42         /**
  43          * Unified CJK blocks.
  44          *
  45          * The same definition of a CJK block must be used for both Collation and
  46          * generateCollationData.php. These blocks are omitted from the first
  47          * letter data, as an optimisation measure and because the default UCA table
  48          * is pretty useless for sorting Chinese text anyway. Japanese and Korean
  49          * blocks are not included here, because they are smaller and more useful.
  50          */
  51         private static $cjkBlocks = [
  52                 [ 0x2E80, 0x2EFF ], // CJK Radicals Supplement
  53                 [ 0x2F00, 0x2FDF ], // Kangxi Radicals
  54                 [ 0x2FF0, 0x2FFF ], // Ideographic Description Characters
  55                 [ 0x3000, 0x303F ], // CJK Symbols and Punctuation
  56                 [ 0x31C0, 0x31EF ], // CJK Strokes
  57                 [ 0x3200, 0x32FF ], // Enclosed CJK Letters and Months
  58                 [ 0x3300, 0x33FF ], // CJK Compatibility
  59                 [ 0x3400, 0x4DBF ], // CJK Unified Ideographs Extension A
  60                 [ 0x4E00, 0x9FFF ], // CJK Unified Ideographs
  61                 [ 0xF900, 0xFAFF ], // CJK Compatibility Ideographs
  62                 [ 0xFE30, 0xFE4F ], // CJK Compatibility Forms
  63                 [ 0x20000, 0x2A6DF ], // CJK Unified Ideographs Extension B
  64                 [ 0x2A700, 0x2B73F ], // CJK Unified Ideographs Extension C
  65                 [ 0x2B740, 0x2B81F ], // CJK Unified Ideographs Extension D
  66                 [ 0x2F800, 0x2FA1F ], // CJK Compatibility Ideographs Supplement
  67         ];
  68
  69         /**
  70          * Additional characters (or character groups) to be considered separate
  71          * letters for given languages, or to be removed from the list of such
  72          * letters (denoted by keys starting with '-').
  73          *
  74          * These are additions to (or subtractions from) the data stored in the
  75          * first-letters-root.ser file (which among others includes full basic latin,
  76          * cyrillic and greek alphabets).
  77          *
  78          * "Separate letter" is a letter that would have a separate heading/section
  79          * for it in a dictionary or a phone book in this language. This data isn't
  80          * used for sorting (the ICU library handles that), only for deciding which
  81          * characters (or character groups) to use as headings.
  82          *
  83          * Initially generated based on the primary level of Unicode collation
  84          * tailorings available at http://developer.mimer.com/charts/tailorings.htm ,
  85          * later modified.
  86          *
  87          * Empty arrays are intended; this signifies that the data for the language is
  88          * available and that there are, in fact, no additional letters to consider.
  89          */
  90         private static $tailoringFirstLetters = [
  91                 // Verified by native speakers
  92                 'be' => [ "Ё" ],
  93                 'be-tarask' => [ "Ё" ],
  94                 'cy' => [ "Ch", "Dd", "Ff", "Ng", "Ll", "Ph", "Rh", "Th" ],
  95                 'en' => [],
  96                 'fa' => [ "آ", "ء", "ه" ],
  97                 'fi' => [ "Å", "Ä", "Ö" ],
  98                 'fr' => [],
  99                 'hu' => [ "Cs", "Dz", "Dzs", "Gy", "Ly", "Ny", "Ö", "Sz", "Ty", "Ü", "Zs" ],
 100                 'is' => [ "Á", "Ð", "É", "Í", "Ó", "Ú", "Ý", "Þ", "Æ", "Ö", "Å" ],
 101                 'it' => [],
 102                 'lv' => [ "Č", "Ģ", "Ķ", "Ļ", "Ņ", "Š", "Ž" ],
 103                 'pl' => [ "Ą", "Ć", "Ę", "Ł", "Ń", "Ó", "Ś", "Ź", "Ż" ],
 104                 'pt' => [],
 105                 'ru' => [],
 106                 'sv' => [ "Å", "Ä", "Ö" ],
 107                 'sv@collation=standard' => [ "Å", "Ä", "Ö" ],
 108                 'uk' => [ "Ґ", "Ь" ],
 109                 'vi' => [ "Ă", "Â", "Đ", "Ê", "Ô", "Ơ", "Ư" ],
 110                 // Not verified, but likely correct
 111                 'af' => [],
 112                 'ast' => [ "Ch", "Ll", "Ñ" ],
 113                 'az' => [ "Ç", "Ə", "Ğ", "İ", "Ö", "Ş", "Ü" ],
 114                 'bg' => [],
 115                 'br' => [ "Ch", "C'h" ],
 116                 'bs' => [ "Č", "Ć", "Dž", "Đ", "Lj", "Nj", "Š", "Ž" ],
 117                 'ca' => [],
 118                 'co' => [],
 119                 'cs' => [ "Č", "Ch", "Ř", "Š", "Ž" ],
 120                 'da' => [ "Æ", "Ø", "Å" ],
 121                 'de' => [],
 122                 'dsb' => [ "Č", "Ć", "Dź", "Ě", "Ch", "Ł", "Ń", "Ŕ", "Š", "Ś", "Ž", "Ź" ],
 123                 'el' => [],
 124                 'eo' => [ "Ĉ", "Ĝ", "Ĥ", "Ĵ", "Ŝ", "Ŭ" ],
 125                 'es' => [ "Ñ" ],
 126                 'et' => [ "Š", "Ž", "Õ", "Ä", "Ö", "Ü", "W" ], // added W for CollationEt (xx-uca-et)
 127                 'eu' => [ "Ñ" ],
 128                 'fo' => [ "Á", "Ð", "Í", "Ó", "Ú", "Ý", "Æ", "Ø", "Å" ],
 129                 'fur' => [ "À", "Á", "Â", "È", "Ì", "Ò", "Ù" ],
 130                 'fy' => [],
 131                 'ga' => [],
 132                 'gd' => [],
 133                 'gl' => [ "Ch", "Ll", "Ñ" ],
 134                 'hr' => [ "Č", "Ć", "Dž", "Đ", "Lj", "Nj", "Š", "Ž" ],
 135                 'hsb' => [ "Č", "Dź", "Ě", "Ch", "Ł", "Ń", "Ř", "Š", "Ć", "Ž" ],
 136                 'kk' => [ "Ү", "І" ],
 137                 'kl' => [ "Æ", "Ø", "Å" ],
 138                 'ku' => [ "Ç", "Ê", "Î", "Ş", "Û" ],
 139                 'ky' => [ "Ё" ],
 140                 'la' => [],
 141                 'lb' => [],
 142                 'lt' => [ "Č", "Š", "Ž" ],
 143                 'mk' => [],
 144                 'mo' => [ "Ă", "Â", "Î", "Ş", "Ţ" ],
 145                 'mt' => [ "Ċ", "Ġ", "Għ", "Ħ", "Ż" ],
 146                 'nl' => [],
 147                 'no' => [ "Æ", "Ø", "Å" ],
 148                 'oc' => [],
 149                 'rm' => [],
 150                 'ro' => [ "Ă", "Â", "Î", "Ş", "Ţ" ],
 151                 'rup' => [ "Ă", "Â", "Î", "Ľ", "Ń", "Ş", "Ţ" ],
 152                 'sco' => [],
 153                 'sk' => [ "Ä", "Č", "Ch", "Ô", "Š", "Ž" ],
 154                 'sl' => [ "Č", "Š", "Ž" ],
 155                 'smn' => [ "Á", "Č", "Đ", "Ŋ", "Š", "Ŧ", "Ž", "Æ", "Ø", "Å", "Ä", "Ö" ],
 156                 'sq' => [ "Ç", "Dh", "Ë", "Gj", "Ll", "Nj", "Rr", "Sh", "Th", "Xh", "Zh" ],
 157                 'sr' => [],
 158                 'tk' => [ "Ç", "Ä", "Ž", "Ň", "Ö", "Ş", "Ü", "Ý" ],
 159                 'tl' => [ "Ñ", "Ng" ],
 160                 'tr' => [ "Ç", "Ğ", "İ", "Ö", "Ş", "Ü" ],
 161                 'tt' => [ "Ә", "Ө", "Ү", "Җ", "Ң", "Һ" ],
 162                 'uz' => [ "Ch", "G'", "Ng", "O'", "Sh" ],
 163         ];
 164
 165         /**
 166          * @since 1.16.3
 167          */
 168         const RECORD_LENGTH = 14;
 169
 170         public function __construct( $locale ) {
 171                 if ( !extension_loaded( 'intl' ) ) {
 172                         throw new MWException( 'An ICU collation was requested, ' .
 173                                 'but the intl extension is not available.' );
 174                 }
 175
 176                 $this->locale = $locale;
 177                 // Drop everything after the '@' in locale's name
 178                 $localeParts = explode( '@', $locale );
 179                 $this->digitTransformLanguage = Language::factory( $locale === 'root' ? 'en' : $localeParts[0] );
 180
 181                 $this->mainCollator = Collator::create( $locale );
 182                 if ( !$this->mainCollator ) {
 183                         throw new MWException( "Invalid ICU locale specified for collation: $locale" );
 184                 }
 185
 186                 $this->primaryCollator = Collator::create( $locale );
 187                 $this->primaryCollator->setStrength( Collator::PRIMARY );
 188         }
 189
 190         public function getSortKey( $string ) {
 191                 // intl extension produces non null-terminated
 192                 // strings. Appending '' fixes it so that it doesn't generate
 193                 // a warning on each access in debug php.
 194                 MediaWiki\suppressWarnings();
 195                 $key = $this->mainCollator->getSortKey( $string ) . '';
 196                 MediaWiki\restoreWarnings();
 197                 return $key;
 198         }
 199
 200         public function getPrimarySortKey( $string ) {
 201                 MediaWiki\suppressWarnings();
 202                 $key = $this->primaryCollator->getSortKey( $string ) . '';
 203                 MediaWiki\restoreWarnings();
 204                 return $key;
 205         }
 206
 207         public function getFirstLetter( $string ) {
 208                 $string = strval( $string );
 209                 if ( $string === '' ) {
 210                         return '';
 211                 }
 212
 213                 // Check for CJK
 214                 $firstChar = mb_substr( $string, 0, 1, 'UTF-8' );
 215                 if ( ord( $firstChar ) > 0x7f && self::isCjk( UtfNormal\Utils::utf8ToCodepoint( $firstChar ) ) ) {
 216                         return $firstChar;
 217                 }
 218
 219                 $sortKey = $this->getPrimarySortKey( $string );
 220
 221                 // Do a binary search to find the correct letter to sort under
 222                 $min = ArrayUtils::findLowerBound(
 223                         [ $this, 'getSortKeyByLetterIndex' ],
 224                         $this->getFirstLetterCount(),
 225                         'strcmp',
 226                         $sortKey );
 227
 228                 if ( $min === false ) {
 229                         // Before the first letter
 230                         return '';
 231                 }
 232                 return $this->getLetterByIndex( $min );
 233         }
 234
 235         /**
 236          * @since 1.16.3
 237          * @return array
 238          */
 239         public function getFirstLetterData() {
 240                 if ( $this->firstLetterData === null ) {
 241                         $cache = ObjectCache::getLocalServerInstance( CACHE_ANYTHING );
 242                         $cacheKey = $cache->makeKey(
 243                                 'first-letters',
 244                                 $this->locale,
 245                                 $this->digitTransformLanguage->getCode(),
 246                                 self::getICUVersion(),
 247                                 self::FIRST_LETTER_VERSION
 248                         );
 249                         $this->firstLetterData = $cache->getWithSetCallback( $cacheKey, $cache::TTL_WEEK, function () {
 250                                 return $this->fetchFirstLetterData();
 251                         } );
 252                 }
 253                 return $this->firstLetterData;
 254         }
 255
 256         /**
 257          * @return array
 258          * @throws MWException
 259          */
 260         private function fetchFirstLetterData() {
 261                 // Generate data from serialized data file
 262                 if ( isset( self::$tailoringFirstLetters[$this->locale] ) ) {
 263                         $letters = wfGetPrecompiledData( 'first-letters-root.ser' );
 264                         // Append additional characters
 265                         $letters = array_merge( $letters, self::$tailoringFirstLetters[$this->locale] );
 266                         // Remove unnecessary ones, if any
 267                         if ( isset( self::$tailoringFirstLetters['-' . $this->locale] ) ) {
 268                                 $letters = array_diff( $letters, self::$tailoringFirstLetters['-' . $this->locale] );
 269                         }
 270                         // Apply digit transforms
 271                         $digits = [ '0', '1', '2', '3', '4', '5', '6', '7', '8', '9' ];
 272                         $letters = array_diff( $letters, $digits );
 273                         foreach ( $digits as $digit ) {
 274                                 $letters[] = $this->digitTransformLanguage->formatNum( $digit, true );
 275                         }
 276                 } else {
 277                         $letters = wfGetPrecompiledData( "first-letters-{$this->locale}.ser" );
 278                         if ( $letters === false ) {
 279                                 throw new MWException( "MediaWiki does not support ICU locale " .
 280                                         "\"{$this->locale}\"" );
 281                         }
 282                 }
 283
 284                 /* Sort the letters.
 285                  *
 286                  * It's impossible to have the precompiled data file properly sorted,
 287                  * because the sort order changes depending on ICU version. If the
 288                  * array is not properly sorted, the binary search will return random
 289                  * results.
 290                  *
 291                  * We also take this opportunity to remove primary collisions.
 292                  */
 293                 $letterMap = [];
 294                 foreach ( $letters as $letter ) {
 295                         $key = $this->getPrimarySortKey( $letter );
 296                         if ( isset( $letterMap[$key] ) ) {
 297                                 // Primary collision
 298                                 // Keep whichever one sorts first in the main collator
 299                                 if ( $this->mainCollator->compare( $letter, $letterMap[$key] ) < 0 ) {
 300                                         $letterMap[$key] = $letter;
 301                                 }
 302                         } else {
 303                                 $letterMap[$key] = $letter;
 304                         }
 305                 }
 306                 ksort( $letterMap, SORT_STRING );
 307
 308                 /* Remove duplicate prefixes. Basically if something has a sortkey
 309                  * which is a prefix of some other sortkey, then it is an
 310                  * expansion and probably should not be considered a section
 311                  * header.
 312                  *
 313                  * For example 'þ' is sometimes sorted as if it is the letters
 314                  * 'th'. Other times it is its own primary element. Another
 315                  * example is '₨'. Sometimes its a currency symbol. Sometimes it
 316                  * is an 'R' followed by an 's'.
 317                  *
 318                  * Additionally an expanded element should always sort directly
 319                  * after its first element due to they way sortkeys work.
 320                  *
 321                  * UCA sortkey elements are of variable length but no collation
 322                  * element should be a prefix of some other element, so I think
 323                  * this is safe. See:
 324                  * - https://ssl.icu-project.org/repos/icu/icuhtml/trunk/design/collation/ICU_collation_design.htm
 325                  * - http://site.icu-project.org/design/collation/uca-weight-allocation
 326                  *
 327                  * Additionally, there is something called primary compression to
 328                  * worry about. Basically, if you have two primary elements that
 329                  * are more than one byte and both start with the same byte then
 330                  * the first byte is dropped on the second primary. Additionally
 331                  * either \x03 or \xFF may be added to mean that the next primary
 332                  * does not start with the first byte of the first primary.
 333                  *
 334                  * This shouldn't matter much, as the first primary is not
 335                  * changed, and that is what we are comparing against.
 336                  *
 337                  * tl;dr: This makes some assumptions about how icu implements
 338                  * collations. It seems incredibly unlikely these assumptions
 339                  * will change, but nonetheless they are assumptions.
 340                  */
 341
 342                 $prev = false;
 343                 $duplicatePrefixes = [];
 344                 foreach ( $letterMap as $key => $value ) {
 345                         // Remove terminator byte. Otherwise the prefix
 346                         // comparison will get hung up on that.
 347                         $trimmedKey = rtrim( $key, "\0" );
 348                         if ( $prev === false || $prev === '' ) {
 349                                 $prev = $trimmedKey;
 350                                 // We don't yet have a collation element
 351                                 // to compare against, so continue.
 352                                 continue;
 353                         }
 354
 355                         // Due to the fact the array is sorted, we only have
 356                         // to compare with the element directly previous
 357                         // to the current element (skipping expansions).
 358                         // An element "X" will always sort directly
 359                         // before "XZ" (Unless we have "XY", but we
 360                         // do not update $prev in that case).
 361                         if ( substr( $trimmedKey, 0, strlen( $prev ) ) === $prev ) {
 362                                 $duplicatePrefixes[] = $key;
 363                                 // If this is an expansion, we don't want to
 364                                 // compare the next element to this element,
 365                                 // but to what is currently $prev
 366                                 continue;
 367                         }
 368                         $prev = $trimmedKey;
 369                 }
 370                 foreach ( $duplicatePrefixes as $badKey ) {
 371                         wfDebug( "Removing '{$letterMap[$badKey]}' from first letters.\n" );
 372                         unset( $letterMap[$badKey] );
 373                         // This code assumes that unsetting does not change sort order.
 374                 }
 375                 $data = [
 376                         'chars' => array_values( $letterMap ),
 377                         'keys' => array_keys( $letterMap ),
 378                 ];
 379
 380                 // Reduce memory usage before caching
 381                 unset( $letterMap );
 382
 383                 return $data;
 384         }
 385
 386         /**
 387          * @since 1.16.3
 388          */
 389         public function getLetterByIndex( $index ) {
 390                 return $this->getFirstLetterData()['chars'][$index];
 391         }
 392
 393         /**
 394          * @since 1.16.3
 395          */
 396         public function getSortKeyByLetterIndex( $index ) {
 397                 return $this->getFirstLetterData()['keys'][$index];
 398         }
 399
 400         /**
 401          * @since 1.16.3
 402          */
 403         public function getFirstLetterCount() {
 404                 return count( $this->getFirstLetterData()['chars'] );
 405         }
 406
 407         /**
 408          * @since 1.16.3
 409          */
 410         public static function isCjk( $codepoint ) {
 411                 foreach ( self::$cjkBlocks as $block ) {
 412                         if ( $codepoint >= $block[0] && $codepoint <= $block[1] ) {
 413                                 return true;
 414                         }
 415                 }
 416                 return false;
 417         }
 418
 419         /**
 420          * Return the version of ICU library used by PHP's intl extension,
 421          * or false when the extension is not installed of the version
 422          * can't be determined.
 423          *
 424          * The constant INTL_ICU_VERSION this function refers to isn't really
 425          * documented. It is available since PHP 5.3.7 (see PHP bug 54561).
 426          * This function will return false on older PHPs.
 427          *
 428          * @since 1.21
 429          * @return string|bool
 430          */
 431         static function getICUVersion() {
 432                 return defined( 'INTL_ICU_VERSION' ) ? INTL_ICU_VERSION : false;
 433         }
 434
 435         /**
 436          * Return the version of Unicode appropriate for the version of ICU library
 437          * currently in use, or false when it can't be determined.
 438          *
 439          * @since 1.21
 440          * @return string|bool
 441          */
 442         static function getUnicodeVersionForICU() {
 443                 $icuVersion = IcuCollation::getICUVersion();
 444                 if ( !$icuVersion ) {
 445                         return false;
 446                 }
 447
 448                 $versionPrefix = substr( $icuVersion, 0, 3 );
 449                 // Source: http://site.icu-project.org/download
 450                 $map = [
 451                         '50.' => '6.2',
 452                         '49.' => '6.1',
 453                         '4.8' => '6.0',
 454                         '4.6' => '6.0',
 455                         '4.4' => '5.2',
 456                         '4.2' => '5.1',
 457                         '4.0' => '5.1',
 458                         '3.8' => '5.0',
 459                         '3.6' => '5.0',
 460                         '3.4' => '4.1',
 461                 ];
 462
 463                 if ( isset( $map[$versionPrefix] ) ) {
 464                         return $map[$versionPrefix];
 465                 } else {
 466                         return false;
 467                 }
 468         }
 469 }