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