includes/Collation.php

   1 <?php
   2 /**
   3  * Database row sorting.
   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  * @file
  21  */
  22
  23 abstract class Collation {
  24         static $instance;
  25
  26         /**
  27          * @return Collation
  28          */
  29         static function singleton() {
  30                 if ( !self::$instance ) {
  31                         global $wgCategoryCollation;
  32                         self::$instance = self::factory( $wgCategoryCollation );
  33                 }
  34                 return self::$instance;
  35         }
  36
  37         /**
  38          * @throws MWException
  39          * @param $collationName string
  40          * @return Collation
  41          */
  42         static function factory( $collationName ) {
  43                 switch ( $collationName ) {
  44                         case 'uppercase':
  45                                 return new UppercaseCollation;
  46                         case 'identity':
  47                                 return new IdentityCollation;
  48                         case 'uca-default':
  49                                 return new IcuCollation( 'root' );
  50                         case 'xx-uca-ckb':
  51                                 return new CollationCkb;
  52                         default:
  53                                 $match = array();
  54                                 if ( preg_match( '/^uca-([a-z@=-]+)$/', $collationName, $match ) ) {
  55                                         return new IcuCollation( $match[1] );
  56                                 }
  57
  58                                 # Provide a mechanism for extensions to hook in.
  59                                 $collationObject = null;
  60                                 wfRunHooks( 'Collation::factory', array( $collationName, &$collationObject ) );
  61
  62                                 if ( $collationObject instanceof Collation ) {
  63                                         return $collationObject;
  64                                 }
  65
  66                                 // If all else fails...
  67                                 throw new MWException( __METHOD__ . ": unknown collation type \"$collationName\"" );
  68                 }
  69         }
  70
  71         /**
  72          * Given a string, convert it to a (hopefully short) key that can be used
  73          * for efficient sorting.  A binary sort according to the sortkeys
  74          * corresponds to a logical sort of the corresponding strings.  Current
  75          * code expects that a line feed character should sort before all others, but
  76          * has no other particular expectations (and that one can be changed if
  77          * necessary).
  78          *
  79          * @param string $string UTF-8 string
  80          * @return string Binary sortkey
  81          */
  82         abstract function getSortKey( $string );
  83
  84         /**
  85          * Given a string, return the logical "first letter" to be used for
  86          * grouping on category pages and so on.  This has to be coordinated
  87          * carefully with convertToSortkey(), or else the sorted list might jump
  88          * back and forth between the same "initial letters" or other pathological
  89          * behavior.  For instance, if you just return the first character, but "a"
  90          * sorts the same as "A" based on getSortKey(), then you might get a
  91          * list like
  92          *
  93          * == A ==
  94          * * [[Aardvark]]
  95          *
  96          * == a ==
  97          * * [[antelope]]
  98          *
  99          * == A ==
 100          * * [[Ape]]
 101          *
 102          * etc., assuming for the sake of argument that $wgCapitalLinks is false.
 103          *
 104          * @param string $string UTF-8 string
 105          * @return string UTF-8 string corresponding to the first letter of input
 106          */
 107         abstract function getFirstLetter( $string );
 108 }
 109
 110 class UppercaseCollation extends Collation {
 111         var $lang;
 112         function __construct() {
 113                 // Get a language object so that we can use the generic UTF-8 uppercase
 114                 // function there
 115                 $this->lang = Language::factory( 'en' );
 116         }
 117
 118         function getSortKey( $string ) {
 119                 return $this->lang->uc( $string );
 120         }
 121
 122         function getFirstLetter( $string ) {
 123                 if ( $string[0] == "\0" ) {
 124                         $string = substr( $string, 1 );
 125                 }
 126                 return $this->lang->ucfirst( $this->lang->firstChar( $string ) );
 127         }
 128 }
 129
 130 /**
 131  * Collation class that's essentially a no-op.
 132  *
 133  * Does sorting based on binary value of the string.
 134  * Like how things were pre 1.17.
 135  */
 136 class IdentityCollation extends Collation {
 137
 138         function getSortKey( $string ) {
 139                 return $string;
 140         }
 141
 142         function getFirstLetter( $string ) {
 143                 global $wgContLang;
 144                 // Copied from UppercaseCollation.
 145                 // I'm kind of unclear on when this could happen...
 146                 if ( $string[0] == "\0" ) {
 147                         $string = substr( $string, 1 );
 148                 }
 149                 return $wgContLang->firstChar( $string );
 150         }
 151 }
 152
 153 class IcuCollation extends Collation {
 154         const FIRST_LETTER_VERSION = 2;
 155
 156         var $primaryCollator, $mainCollator, $locale, $digitTransformLanguage;
 157         var $firstLetterData;
 158
 159         /**
 160          * Unified CJK blocks.
 161          *
 162          * The same definition of a CJK block must be used for both Collation and
 163          * generateCollationData.php. These blocks are omitted from the first
 164          * letter data, as an optimisation measure and because the default UCA table
 165          * is pretty useless for sorting Chinese text anyway. Japanese and Korean
 166          * blocks are not included here, because they are smaller and more useful.
 167          */
 168         static $cjkBlocks = array(
 169                 array( 0x2E80, 0x2EFF ), // CJK Radicals Supplement
 170                 array( 0x2F00, 0x2FDF ), // Kangxi Radicals
 171                 array( 0x2FF0, 0x2FFF ), // Ideographic Description Characters
 172                 array( 0x3000, 0x303F ), // CJK Symbols and Punctuation
 173                 array( 0x31C0, 0x31EF ), // CJK Strokes
 174                 array( 0x3200, 0x32FF ), // Enclosed CJK Letters and Months
 175                 array( 0x3300, 0x33FF ), // CJK Compatibility
 176                 array( 0x3400, 0x4DBF ), // CJK Unified Ideographs Extension A
 177                 array( 0x4E00, 0x9FFF ), // CJK Unified Ideographs
 178                 array( 0xF900, 0xFAFF ), // CJK Compatibility Ideographs
 179                 array( 0xFE30, 0xFE4F ), // CJK Compatibility Forms
 180                 array( 0x20000, 0x2A6DF ), // CJK Unified Ideographs Extension B
 181                 array( 0x2A700, 0x2B73F ), // CJK Unified Ideographs Extension C
 182                 array( 0x2B740, 0x2B81F ), // CJK Unified Ideographs Extension D
 183                 array( 0x2F800, 0x2FA1F ), // CJK Compatibility Ideographs Supplement
 184         );
 185
 186         /**
 187          * Additional characters (or character groups) to be considered separate
 188          * letters for given languages, or to be removed from the list of such
 189          * letters (denoted by keys starting with '-').
 190          *
 191          * These are additions to (or subtractions from) the data stored in the
 192          * first-letters-root.ser file (which among others includes full basic latin,
 193          * cyrillic and greek alphabets).
 194          *
 195          * "Separate letter" is a letter that would have a separate heading/section
 196          * for it in a dictionary or a phone book in this language. This data isn't
 197          * used for sorting (the ICU library handles that), only for deciding which
 198          * characters (or character groups) to use as headings.
 199          *
 200          * Initially generated based on the primary level of Unicode collation
 201          * tailorings available at http://developer.mimer.com/charts/tailorings.htm ,
 202          * later modified.
 203          *
 204          * Empty arrays are intended; this signifies that the data for the language is
 205          * available and that there are, in fact, no additional letters to consider.
 206          */
 207         static $tailoringFirstLetters = array(
 208                 // Verified by native speakers
 209                 'be' => array( "Ё" ),
 210                 'be-tarask' => array( "Ё" ),
 211                 'en' => array(),
 212                 'fi' => array( "Å", "Ä", "Ö" ),
 213                 'hu' => array( "Cs", "Dz", "Dzs", "Gy", "Ly", "Ny", "Ö", "Sz", "Ty", "Ü", "Zs" ),
 214                 'it' => array(),
 215                 'pl' => array( "Ą", "Ć", "Ę", "Ł", "Ń", "Ó", "Ś", "Ź", "Ż" ),
 216                 'pt' => array(),
 217                 'ru' => array(),
 218                 'sv' => array( "Å", "Ä", "Ö" ),
 219                 'sv@collation=standard' => array( "Å", "Ä", "Ö" ),
 220                 'uk' => array( "Ґ", "Ь" ),
 221                 'vi' => array( "Ă", "Â", "Đ", "Ê", "Ô", "Ơ", "Ư" ),
 222                 // Not verified, but likely correct
 223                 'af' => array(),
 224                 'ast' => array( "Ch", "Ll", "Ñ" ),
 225                 'az' => array( "Ç", "Ə", "Ğ", "İ", "Ö", "Ş", "Ü" ),
 226                 'bg' => array(),
 227                 'br' => array( "Ch", "C'h" ),
 228                 'bs' => array( "Č", "Ć", "Dž", "Đ", "Lj", "Nj", "Š", "Ž" ),
 229                 'ca' => array(),
 230                 'co' => array(),
 231                 'cs' => array( "Č", "Ch", "Ř", "Š", "Ž" ),
 232                 'cy' => array( "Ch", "Dd", "Ff", "Ng", "Ll", "Ph", "Rh", "Th" ),
 233                 'da' => array( "Æ", "Ø", "Å" ),
 234                 'de' => array(),
 235                 'dsb' => array( "Č", "Ć", "Dź", "Ě", "Ch", "Ł", "Ń", "Ŕ", "Š", "Ś", "Ž", "Ź" ),
 236                 'el' => array(),
 237                 'eo' => array( "Ĉ", "Ĝ", "Ĥ", "Ĵ", "Ŝ", "Ŭ" ),
 238                 'es' => array( "Ñ" ),
 239                 'et' => array( "Š", "Ž", "Õ", "Ä", "Ö", "Ü" ),
 240                 'eu' => array( "Ñ" ),
 241                 'fa' => array( "آ", "ء", "ه" ),
 242                 'fo' => array( "Á", "Ð", "Í", "Ó", "Ú", "Ý", "Æ", "Ø", "Å" ),
 243                 'fr' => array(),
 244                 'fur' => array( "À", "Á", "Â", "È", "Ì", "Ò", "Ù" ),
 245                 'fy' => array(),
 246                 'ga' => array(),
 247                 'gd' => array(),
 248                 'gl' => array( "Ch", "Ll", "Ñ" ),
 249                 'hr' => array( "Č", "Ć", "Dž", "Đ", "Lj", "Nj", "Š", "Ž" ),
 250                 'hsb' => array( "Č", "Dź", "Ě", "Ch", "Ł", "Ń", "Ř", "Š", "Ć", "Ž" ),
 251                 'is' => array( "Á", "Ð", "É", "Í", "Ó", "Ú", "Ý", "Þ", "Æ", "Ö", "Å" ),
 252                 'kk' => array( "Ү", "І" ),
 253                 'kl' => array( "Æ", "Ø", "Å" ),
 254                 'ku' => array( "Ç", "Ê", "Î", "Ş", "Û" ),
 255                 'ky' => array( "Ё" ),
 256                 'la' => array(),
 257                 'lb' => array(),
 258                 'lt' => array( "Č", "Š", "Ž" ),
 259                 'lv' => array( "Č", "Ģ", "Ķ", "Ļ", "Ņ", "Š", "Ž" ),
 260                 'mk' => array(),
 261                 'mo' => array( "Ă", "Â", "Î", "Ş", "Ţ" ),
 262                 'mt' => array( "Ċ", "Ġ", "Għ", "Ħ", "Ż" ),
 263                 'nl' => array(),
 264                 'no' => array( "Æ", "Ø", "Å" ),
 265                 'oc' => array(),
 266                 'rm' => array(),
 267                 'ro' => array( "Ă", "Â", "Î", "Ş", "Ţ" ),
 268                 'rup' => array( "Ă", "Â", "Î", "Ľ", "Ń", "Ş", "Ţ" ),
 269                 'sco' => array(),
 270                 'sk' => array( "Ä", "Č", "Ch", "Ô", "Š", "Ž" ),
 271                 'sl' => array( "Č", "Š", "Ž" ),
 272                 'smn' => array( "Á", "Č", "Đ", "Ŋ", "Š", "Ŧ", "Ž", "Æ", "Ø", "Å", "Ä", "Ö" ),
 273                 'sq' => array( "Ç", "Dh", "Ë", "Gj", "Ll", "Nj", "Rr", "Sh", "Th", "Xh", "Zh" ),
 274                 'sr' => array(),
 275                 'tk' => array( "Ç", "Ä", "Ž", "Ň", "Ö", "Ş", "Ü", "Ý" ),
 276                 'tl' => array( "Ñ", "Ng" ),
 277                 'tr' => array( "Ç", "Ğ", "İ", "Ö", "Ş", "Ü" ),
 278                 'tt' => array( "Ә", "Ө", "Ү", "Җ", "Ң", "Һ" ),
 279                 'uz' => array( "Ch", "G'", "Ng", "O'", "Sh" ),
 280         );
 281
 282         const RECORD_LENGTH = 14;
 283
 284         function __construct( $locale ) {
 285                 if ( !extension_loaded( 'intl' ) ) {
 286                         throw new MWException( 'An ICU collation was requested, ' .
 287                                 'but the intl extension is not available.' );
 288                 }
 289
 290                 $this->locale = $locale;
 291                 // Drop everything after the '@' in locale's name
 292                 $localeParts = explode( '@', $locale );
 293                 $this->digitTransformLanguage = Language::factory( $locale === 'root' ? 'en' : $localeParts[0] );
 294
 295                 $this->mainCollator = Collator::create( $locale );
 296                 if ( !$this->mainCollator ) {
 297                         throw new MWException( "Invalid ICU locale specified for collation: $locale" );
 298                 }
 299
 300                 $this->primaryCollator = Collator::create( $locale );
 301                 $this->primaryCollator->setStrength( Collator::PRIMARY );
 302         }
 303
 304         function getSortKey( $string ) {
 305                 // intl extension produces non null-terminated
 306                 // strings. Appending '' fixes it so that it doesn't generate
 307                 // a warning on each access in debug php.
 308                 wfSuppressWarnings();
 309                 $key = $this->mainCollator->getSortKey( $string ) . '';
 310                 wfRestoreWarnings();
 311                 return $key;
 312         }
 313
 314         function getPrimarySortKey( $string ) {
 315                 wfSuppressWarnings();
 316                 $key = $this->primaryCollator->getSortKey( $string ) . '';
 317                 wfRestoreWarnings();
 318                 return $key;
 319         }
 320
 321         function getFirstLetter( $string ) {
 322                 $string = strval( $string );
 323                 if ( $string === '' ) {
 324                         return '';
 325                 }
 326
 327                 // Check for CJK
 328                 $firstChar = mb_substr( $string, 0, 1, 'UTF-8' );
 329                 if ( ord( $firstChar ) > 0x7f && self::isCjk( utf8ToCodepoint( $firstChar ) ) ) {
 330                         return $firstChar;
 331                 }
 332
 333                 $sortKey = $this->getPrimarySortKey( $string );
 334
 335                 // Do a binary search to find the correct letter to sort under
 336                 $min = ArrayUtils::findLowerBound(
 337                         array( $this, 'getSortKeyByLetterIndex' ),
 338                         $this->getFirstLetterCount(),
 339                         'strcmp',
 340                         $sortKey );
 341
 342                 if ( $min === false ) {
 343                         // Before the first letter
 344                         return '';
 345                 }
 346                 return $this->getLetterByIndex( $min );
 347         }
 348
 349         function getFirstLetterData() {
 350                 if ( $this->firstLetterData !== null ) {
 351                         return $this->firstLetterData;
 352                 }
 353
 354                 $cache = wfGetCache( CACHE_ANYTHING );
 355                 $cacheKey = wfMemcKey( 'first-letters', $this->locale, $this->digitTransformLanguage->getCode() );
 356                 $cacheEntry = $cache->get( $cacheKey );
 357
 358                 if ( $cacheEntry && isset( $cacheEntry['version'] )
 359                         && $cacheEntry['version'] == self::FIRST_LETTER_VERSION
 360                 ) {
 361                         $this->firstLetterData = $cacheEntry;
 362                         return $this->firstLetterData;
 363                 }
 364
 365                 // Generate data from serialized data file
 366
 367                 if ( isset( self::$tailoringFirstLetters[$this->locale] ) ) {
 368                         $letters = wfGetPrecompiledData( "first-letters-root.ser" );
 369                         // Append additional characters
 370                         $letters = array_merge( $letters, self::$tailoringFirstLetters[$this->locale] );
 371                         // Remove unnecessary ones, if any
 372                         if ( isset( self::$tailoringFirstLetters['-' . $this->locale] ) ) {
 373                                 $letters = array_diff( $letters, self::$tailoringFirstLetters['-' . $this->locale] );
 374                         }
 375                         // Apply digit transforms
 376                         $digits = array( '0', '1', '2', '3', '4', '5', '6', '7', '8', '9' );
 377                         $letters = array_diff( $letters, $digits );
 378                         foreach ( $digits as $digit ) {
 379                                 $letters[] = $this->digitTransformLanguage->formatNum( $digit, true );
 380                         }
 381                 } else {
 382                         $letters = wfGetPrecompiledData( "first-letters-{$this->locale}.ser" );
 383                         if ( $letters === false ) {
 384                                 throw new MWException( "MediaWiki does not support ICU locale " .
 385                                         "\"{$this->locale}\"" );
 386                         }
 387                 }
 388
 389                 // Sort the letters.
 390                 //
 391                 // It's impossible to have the precompiled data file properly sorted,
 392                 // because the sort order changes depending on ICU version. If the
 393                 // array is not properly sorted, the binary search will return random
 394                 // results.
 395                 //
 396                 // We also take this opportunity to remove primary collisions.
 397                 $letterMap = array();
 398                 foreach ( $letters as $letter ) {
 399                         $key = $this->getPrimarySortKey( $letter );
 400                         if ( isset( $letterMap[$key] ) ) {
 401                                 // Primary collision
 402                                 // Keep whichever one sorts first in the main collator
 403                                 if ( $this->mainCollator->compare( $letter, $letterMap[$key] ) < 0 ) {
 404                                         $letterMap[$key] = $letter;
 405                                 }
 406                         } else {
 407                                 $letterMap[$key] = $letter;
 408                         }
 409                 }
 410                 ksort( $letterMap, SORT_STRING );
 411                 // Remove duplicate prefixes. Basically if something has a sortkey
 412                 // which is a prefix of some other sortkey, then it is an
 413                 // expansion and probably should not be considered a section
 414                 // header.
 415                 //
 416                 // For example 'þ' is sometimes sorted as if it is the letters
 417                 // 'th'. Other times it is its own primary element. Another
 418                 // example is '₨'. Sometimes its a currency symbol. Sometimes it
 419                 // is an 'R' followed by an 's'.
 420                 //
 421                 // Additionally an expanded element should always sort directly
 422                 // after its first element due to they way sortkeys work.
 423                 //
 424                 // UCA sortkey elements are of variable length but no collation
 425                 // element should be a prefix of some other element, so I think
 426                 // this is safe. See:
 427                 // * https://ssl.icu-project.org/repos/icu/icuhtml/trunk/design/collation/ICU_collation_design.htm
 428                 // * http://site.icu-project.org/design/collation/uca-weight-allocation
 429                 //
 430                 // Additionally, there is something called primary compression to
 431                 // worry about. Basically, if you have two primary elements that
 432                 // are more than one byte and both start with the same byte then
 433                 // the first byte is dropped on the second primary. Additionally
 434                 // either \x03 or \xFF may be added to mean that the next primary
 435                 // does not start with the first byte of the first primary.
 436                 //
 437                 // This shouldn't matter much, as the first primary is not
 438                 // changed, and that is what we are comparing against.
 439                 //
 440                 // tl;dr: This makes some assumptions about how icu implements
 441                 // collations. It seems incredibly unlikely these assumptions
 442                 // will change, but nonetheless they are assumptions.
 443
 444                 $prev = false;
 445                 $duplicatePrefixes = array();
 446                 foreach ( $letterMap as $key => $value ) {
 447                         // Remove terminator byte. Otherwise the prefix
 448                         // comparison will get hung up on that.
 449                         $trimmedKey = rtrim( $key, "\0" );
 450                         if ( $prev === false || $prev === '' ) {
 451                                 $prev = $trimmedKey;
 452                                 // We don't yet have a collation element
 453                                 // to compare against, so continue.
 454                                 continue;
 455                         }
 456
 457                         // Due to the fact the array is sorted, we only have
 458                         // to compare with the element directly previous
 459                         // to the current element (skipping expansions).
 460                         // An element "X" will always sort directly
 461                         // before "XZ" (Unless we have "XY", but we
 462                         // do not update $prev in that case).
 463                         if ( substr( $trimmedKey, 0, strlen( $prev ) ) === $prev ) {
 464                                 $duplicatePrefixes[] = $key;
 465                                 // If this is an expansion, we don't want to
 466                                 // compare the next element to this element,
 467                                 // but to what is currently $prev
 468                                 continue;
 469                         }
 470                         $prev = $trimmedKey;
 471                 }
 472                 foreach ( $duplicatePrefixes as $badKey ) {
 473                         wfDebug( "Removing '{$letterMap[$badKey]}' from first letters." );
 474                         unset( $letterMap[$badKey] );
 475                         // This code assumes that unsetting does not change sort order.
 476                 }
 477                 $data = array(
 478                         'chars' => array_values( $letterMap ),
 479                         'keys' => array_keys( $letterMap ),
 480                         'version' => self::FIRST_LETTER_VERSION,
 481                 );
 482
 483                 // Reduce memory usage before caching
 484                 unset( $letterMap );
 485
 486                 // Save to cache
 487                 $this->firstLetterData = $data;
 488                 $cache->set( $cacheKey, $data, 86400 * 7 /* 1 week */ );
 489                 return $data;
 490         }
 491
 492         function getLetterByIndex( $index ) {
 493                 if ( $this->firstLetterData === null ) {
 494                         $this->getFirstLetterData();
 495                 }
 496                 return $this->firstLetterData['chars'][$index];
 497         }
 498
 499         function getSortKeyByLetterIndex( $index ) {
 500                 if ( $this->firstLetterData === null ) {
 501                         $this->getFirstLetterData();
 502                 }
 503                 return $this->firstLetterData['keys'][$index];
 504         }
 505
 506         function getFirstLetterCount() {
 507                 if ( $this->firstLetterData === null ) {
 508                         $this->getFirstLetterData();
 509                 }
 510                 return count( $this->firstLetterData['chars'] );
 511         }
 512
 513         /**
 514          * Do a binary search, and return the index of the largest item that sorts
 515          * less than or equal to the target value.
 516          *
 517          * @deprecated in 1.23; use ArrayUtils::findLowerBound() instead
 518          *
 519          * @param array $valueCallback A function to call to get the value with
 520          *     a given array index.
 521          * @param int $valueCount The number of items accessible via $valueCallback,
 522          *     indexed from 0 to $valueCount - 1
 523          * @param array $comparisonCallback A callback to compare two values, returning
 524          *     -1, 0 or 1 in the style of strcmp().
 525          * @param string $target The target value to find.
 526          *
 527          * @return int|bool The item index of the lower bound, or false if the target value
 528          *     sorts before all items.
 529          */
 530         function findLowerBound( $valueCallback, $valueCount, $comparisonCallback, $target ) {
 531                 wfDeprecated( __METHOD__, '1.23' );
 532                 return ArrayUtils::findLowerBound( $valueCallback, $valueCount, $comparisonCallback, $target );
 533         }
 534
 535         static function isCjk( $codepoint ) {
 536                 foreach ( self::$cjkBlocks as $block ) {
 537                         if ( $codepoint >= $block[0] && $codepoint <= $block[1] ) {
 538                                 return true;
 539                         }
 540                 }
 541                 return false;
 542         }
 543
 544         /**
 545          * Return the version of ICU library used by PHP's intl extension,
 546          * or false when the extension is not installed of the version
 547          * can't be determined.
 548          *
 549          * The constant INTL_ICU_VERSION this function refers to isn't really
 550          * documented. It is available since PHP 5.3.7 (see PHP bug 54561).
 551          * This function will return false on older PHPs.
 552          *
 553          * @since 1.21
 554          * @return string|false
 555          */
 556         static function getICUVersion() {
 557                 return defined( 'INTL_ICU_VERSION' ) ? INTL_ICU_VERSION : false;
 558         }
 559
 560         /**
 561          * Return the version of Unicode appropriate for the version of ICU library
 562          * currently in use, or false when it can't be determined.
 563          *
 564          * @since 1.21
 565          * @return string|false
 566          */
 567         static function getUnicodeVersionForICU() {
 568                 $icuVersion = IcuCollation::getICUVersion();
 569                 if ( !$icuVersion ) {
 570                         return false;
 571                 }
 572
 573                 $versionPrefix = substr( $icuVersion, 0, 3 );
 574                 // Source: http://site.icu-project.org/download
 575                 $map = array(
 576                         '50.' => '6.2',
 577                         '49.' => '6.1',
 578                         '4.8' => '6.0',
 579                         '4.6' => '6.0',
 580                         '4.4' => '5.2',
 581                         '4.2' => '5.1',
 582                         '4.0' => '5.1',
 583                         '3.8' => '5.0',
 584                         '3.6' => '5.0',
 585                         '3.4' => '4.1',
 586                 );
 587
 588                 if ( isset( $map[$versionPrefix] ) ) {
 589                         return $map[$versionPrefix];
 590                 } else {
 591                         return false;
 592                 }
 593         }
 594 }
 595
 596 /**
 597  * Workaround for the lack of support of Sorani Kurdish / Central Kurdish language ('ckb') in ICU.
 598  *
 599  * Uses the same collation rules as Persian / Farsi ('fa'), but different characters for digits.
 600  */
 601 class CollationCkb extends IcuCollation {
 602         function __construct() {
 603                 // This will set $locale and collators, which affect the actual sorting order
 604                 parent::__construct( 'fa' );
 605                 // Override the 'fa' language set by parent constructor, which affects #getFirstLetterData()
 606                 $this->digitTransformLanguage = Language::factory( 'ckb' );
 607         }
 608 }