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         private 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 string $collationName
  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         private $lang;
 112
 113         function __construct() {
 114                 // Get a language object so that we can use the generic UTF-8 uppercase
 115                 // function there
 116                 $this->lang = Language::factory( 'en' );
 117         }
 118
 119         function getSortKey( $string ) {
 120                 return $this->lang->uc( $string );
 121         }
 122
 123         function getFirstLetter( $string ) {
 124                 if ( $string[0] == "\0" ) {
 125                         $string = substr( $string, 1 );
 126                 }
 127                 return $this->lang->ucfirst( $this->lang->firstChar( $string ) );
 128         }
 129 }
 130
 131 /**
 132  * Collation class that's essentially a no-op.
 133  *
 134  * Does sorting based on binary value of the string.
 135  * Like how things were pre 1.17.
 136  */
 137 class IdentityCollation extends Collation {
 138
 139         function getSortKey( $string ) {
 140                 return $string;
 141         }
 142
 143         function getFirstLetter( $string ) {
 144                 global $wgContLang;
 145                 // Copied from UppercaseCollation.
 146                 // I'm kind of unclear on when this could happen...
 147                 if ( $string[0] == "\0" ) {
 148                         $string = substr( $string, 1 );
 149                 }
 150                 return $wgContLang->firstChar( $string );
 151         }
 152 }
 153
 154 class IcuCollation extends Collation {
 155         const FIRST_LETTER_VERSION = 2;
 156
 157         /** @var Collator */
 158         private $primaryCollator;
 159
 160         /** @var Collator */
 161         private $mainCollator;
 162
 163         /** @var  */
 164         private $locale;
 165
 166         /** @var Language */
 167         protected $digitTransformLanguage;
 168
 169         /** @var array */
 170         private $firstLetterData;
 171
 172         /**
 173          * Unified CJK blocks.
 174          *
 175          * The same definition of a CJK block must be used for both Collation and
 176          * generateCollationData.php. These blocks are omitted from the first
 177          * letter data, as an optimisation measure and because the default UCA table
 178          * is pretty useless for sorting Chinese text anyway. Japanese and Korean
 179          * blocks are not included here, because they are smaller and more useful.
 180          */
 181         private static $cjkBlocks = array(
 182                 array( 0x2E80, 0x2EFF ), // CJK Radicals Supplement
 183                 array( 0x2F00, 0x2FDF ), // Kangxi Radicals
 184                 array( 0x2FF0, 0x2FFF ), // Ideographic Description Characters
 185                 array( 0x3000, 0x303F ), // CJK Symbols and Punctuation
 186                 array( 0x31C0, 0x31EF ), // CJK Strokes
 187                 array( 0x3200, 0x32FF ), // Enclosed CJK Letters and Months
 188                 array( 0x3300, 0x33FF ), // CJK Compatibility
 189                 array( 0x3400, 0x4DBF ), // CJK Unified Ideographs Extension A
 190                 array( 0x4E00, 0x9FFF ), // CJK Unified Ideographs
 191                 array( 0xF900, 0xFAFF ), // CJK Compatibility Ideographs
 192                 array( 0xFE30, 0xFE4F ), // CJK Compatibility Forms
 193                 array( 0x20000, 0x2A6DF ), // CJK Unified Ideographs Extension B
 194                 array( 0x2A700, 0x2B73F ), // CJK Unified Ideographs Extension C
 195                 array( 0x2B740, 0x2B81F ), // CJK Unified Ideographs Extension D
 196                 array( 0x2F800, 0x2FA1F ), // CJK Compatibility Ideographs Supplement
 197         );
 198
 199         /**
 200          * Additional characters (or character groups) to be considered separate
 201          * letters for given languages, or to be removed from the list of such
 202          * letters (denoted by keys starting with '-').
 203          *
 204          * These are additions to (or subtractions from) the data stored in the
 205          * first-letters-root.ser file (which among others includes full basic latin,
 206          * cyrillic and greek alphabets).
 207          *
 208          * "Separate letter" is a letter that would have a separate heading/section
 209          * for it in a dictionary or a phone book in this language. This data isn't
 210          * used for sorting (the ICU library handles that), only for deciding which
 211          * characters (or character groups) to use as headings.
 212          *
 213          * Initially generated based on the primary level of Unicode collation
 214          * tailorings available at http://developer.mimer.com/charts/tailorings.htm ,
 215          * later modified.
 216          *
 217          * Empty arrays are intended; this signifies that the data for the language is
 218          * available and that there are, in fact, no additional letters to consider.
 219          */
 220         private static $tailoringFirstLetters = array(
 221                 // Verified by native speakers
 222                 'be' => array( "Ё" ),
 223                 'be-tarask' => array( "Ё" ),
 224                 'cy' => array( "Ch", "Dd", "Ff", "Ng", "Ll", "Ph", "Rh", "Th" ),
 225                 'en' => array(),
 226                 'fa' => array( "آ", "ء", "ه" ),
 227                 'fi' => array( "Å", "Ä", "Ö" ),
 228                 'fr' => array(),
 229                 'hu' => array( "Cs", "Dz", "Dzs", "Gy", "Ly", "Ny", "Ö", "Sz", "Ty", "Ü", "Zs" ),
 230                 'is' => array( "Á", "Ð", "É", "Í", "Ó", "Ú", "Ý", "Þ", "Æ", "Ö", "Å" ),
 231                 'it' => array(),
 232                 'lv' => array( "Č", "Ģ", "Ķ", "Ļ", "Ņ", "Š", "Ž" ),
 233                 'pl' => array( "Ą", "Ć", "Ę", "Ł", "Ń", "Ó", "Ś", "Ź", "Ż" ),
 234                 'pt' => array(),
 235                 'ru' => array(),
 236                 'sv' => array( "Å", "Ä", "Ö" ),
 237                 'sv@collation=standard' => array( "Å", "Ä", "Ö" ),
 238                 'uk' => array( "Ґ", "Ь" ),
 239                 'vi' => array( "Ă", "Â", "Đ", "Ê", "Ô", "Ơ", "Ư" ),
 240                 // Not verified, but likely correct
 241                 'af' => array(),
 242                 'ast' => array( "Ch", "Ll", "Ñ" ),
 243                 'az' => array( "Ç", "Ə", "Ğ", "İ", "Ö", "Ş", "Ü" ),
 244                 'bg' => array(),
 245                 'br' => array( "Ch", "C'h" ),
 246                 'bs' => array( "Č", "Ć", "Dž", "Đ", "Lj", "Nj", "Š", "Ž" ),
 247                 'ca' => array(),
 248                 'co' => array(),
 249                 'cs' => array( "Č", "Ch", "Ř", "Š", "Ž" ),
 250                 'da' => array( "Æ", "Ø", "Å" ),
 251                 'de' => array(),
 252                 'dsb' => array( "Č", "Ć", "Dź", "Ě", "Ch", "Ł", "Ń", "Ŕ", "Š", "Ś", "Ž", "Ź" ),
 253                 'el' => array(),
 254                 'eo' => array( "Ĉ", "Ĝ", "Ĥ", "Ĵ", "Ŝ", "Ŭ" ),
 255                 'es' => array( "Ñ" ),
 256                 'et' => array( "Š", "Ž", "Õ", "Ä", "Ö", "Ü" ),
 257                 'eu' => array( "Ñ" ),
 258                 'fo' => array( "Á", "Ð", "Í", "Ó", "Ú", "Ý", "Æ", "Ø", "Å" ),
 259                 'fur' => array( "À", "Á", "Â", "È", "Ì", "Ò", "Ù" ),
 260                 'fy' => array(),
 261                 'ga' => array(),
 262                 'gd' => array(),
 263                 'gl' => array( "Ch", "Ll", "Ñ" ),
 264                 'hr' => array( "Č", "Ć", "Dž", "Đ", "Lj", "Nj", "Š", "Ž" ),
 265                 'hsb' => array( "Č", "Dź", "Ě", "Ch", "Ł", "Ń", "Ř", "Š", "Ć", "Ž" ),
 266                 'kk' => array( "Ү", "І" ),
 267                 'kl' => array( "Æ", "Ø", "Å" ),
 268                 'ku' => array( "Ç", "Ê", "Î", "Ş", "Û" ),
 269                 'ky' => array( "Ё" ),
 270                 'la' => array(),
 271                 'lb' => array(),
 272                 'lt' => array( "Č", "Š", "Ž" ),
 273                 'mk' => array(),
 274                 'mo' => array( "Ă", "Â", "Î", "Ş", "Ţ" ),
 275                 'mt' => array( "Ċ", "Ġ", "Għ", "Ħ", "Ż" ),
 276                 'nl' => array(),
 277                 'no' => array( "Æ", "Ø", "Å" ),
 278                 'oc' => array(),
 279                 'rm' => array(),
 280                 'ro' => array( "Ă", "Â", "Î", "Ş", "Ţ" ),
 281                 'rup' => array( "Ă", "Â", "Î", "Ľ", "Ń", "Ş", "Ţ" ),
 282                 'sco' => array(),
 283                 'sk' => array( "Ä", "Č", "Ch", "Ô", "Š", "Ž" ),
 284                 'sl' => array( "Č", "Š", "Ž" ),
 285                 'smn' => array( "Á", "Č", "Đ", "Ŋ", "Š", "Ŧ", "Ž", "Æ", "Ø", "Å", "Ä", "Ö" ),
 286                 'sq' => array( "Ç", "Dh", "Ë", "Gj", "Ll", "Nj", "Rr", "Sh", "Th", "Xh", "Zh" ),
 287                 'sr' => array(),
 288                 'tk' => array( "Ç", "Ä", "Ž", "Ň", "Ö", "Ş", "Ü", "Ý" ),
 289                 'tl' => array( "Ñ", "Ng" ),
 290                 'tr' => array( "Ç", "Ğ", "İ", "Ö", "Ş", "Ü" ),
 291                 'tt' => array( "Ә", "Ө", "Ү", "Җ", "Ң", "Һ" ),
 292                 'uz' => array( "Ch", "G'", "Ng", "O'", "Sh" ),
 293         );
 294
 295         const RECORD_LENGTH = 14;
 296
 297         function __construct( $locale ) {
 298                 if ( !extension_loaded( 'intl' ) ) {
 299                         throw new MWException( 'An ICU collation was requested, ' .
 300                                 'but the intl extension is not available.' );
 301                 }
 302
 303                 $this->locale = $locale;
 304                 // Drop everything after the '@' in locale's name
 305                 $localeParts = explode( '@', $locale );
 306                 $this->digitTransformLanguage = Language::factory( $locale === 'root' ? 'en' : $localeParts[0] );
 307
 308                 $this->mainCollator = Collator::create( $locale );
 309                 if ( !$this->mainCollator ) {
 310                         throw new MWException( "Invalid ICU locale specified for collation: $locale" );
 311                 }
 312
 313                 $this->primaryCollator = Collator::create( $locale );
 314                 $this->primaryCollator->setStrength( Collator::PRIMARY );
 315         }
 316
 317         function getSortKey( $string ) {
 318                 // intl extension produces non null-terminated
 319                 // strings. Appending '' fixes it so that it doesn't generate
 320                 // a warning on each access in debug php.
 321                 wfSuppressWarnings();
 322                 $key = $this->mainCollator->getSortKey( $string ) . '';
 323                 wfRestoreWarnings();
 324                 return $key;
 325         }
 326
 327         function getPrimarySortKey( $string ) {
 328                 wfSuppressWarnings();
 329                 $key = $this->primaryCollator->getSortKey( $string ) . '';
 330                 wfRestoreWarnings();
 331                 return $key;
 332         }
 333
 334         function getFirstLetter( $string ) {
 335                 $string = strval( $string );
 336                 if ( $string === '' ) {
 337                         return '';
 338                 }
 339
 340                 // Check for CJK
 341                 $firstChar = mb_substr( $string, 0, 1, 'UTF-8' );
 342                 if ( ord( $firstChar ) > 0x7f && self::isCjk( utf8ToCodepoint( $firstChar ) ) ) {
 343                         return $firstChar;
 344                 }
 345
 346                 $sortKey = $this->getPrimarySortKey( $string );
 347
 348                 // Do a binary search to find the correct letter to sort under
 349                 $min = ArrayUtils::findLowerBound(
 350                         array( $this, 'getSortKeyByLetterIndex' ),
 351                         $this->getFirstLetterCount(),
 352                         'strcmp',
 353                         $sortKey );
 354
 355                 if ( $min === false ) {
 356                         // Before the first letter
 357                         return '';
 358                 }
 359                 return $this->getLetterByIndex( $min );
 360         }
 361
 362         function getFirstLetterData() {
 363                 if ( $this->firstLetterData !== null ) {
 364                         return $this->firstLetterData;
 365                 }
 366
 367                 $cache = wfGetCache( CACHE_ANYTHING );
 368                 $cacheKey = wfMemcKey( 'first-letters', $this->locale, $this->digitTransformLanguage->getCode() );
 369                 $cacheEntry = $cache->get( $cacheKey );
 370
 371                 if ( $cacheEntry && isset( $cacheEntry['version'] )
 372                         && $cacheEntry['version'] == self::FIRST_LETTER_VERSION
 373                 ) {
 374                         $this->firstLetterData = $cacheEntry;
 375                         return $this->firstLetterData;
 376                 }
 377
 378                 // Generate data from serialized data file
 379
 380                 if ( isset( self::$tailoringFirstLetters[$this->locale] ) ) {
 381                         $letters = wfGetPrecompiledData( "first-letters-root.ser" );
 382                         // Append additional characters
 383                         $letters = array_merge( $letters, self::$tailoringFirstLetters[$this->locale] );
 384                         // Remove unnecessary ones, if any
 385                         if ( isset( self::$tailoringFirstLetters['-' . $this->locale] ) ) {
 386                                 $letters = array_diff( $letters, self::$tailoringFirstLetters['-' . $this->locale] );
 387                         }
 388                         // Apply digit transforms
 389                         $digits = array( '0', '1', '2', '3', '4', '5', '6', '7', '8', '9' );
 390                         $letters = array_diff( $letters, $digits );
 391                         foreach ( $digits as $digit ) {
 392                                 $letters[] = $this->digitTransformLanguage->formatNum( $digit, true );
 393                         }
 394                 } else {
 395                         $letters = wfGetPrecompiledData( "first-letters-{$this->locale}.ser" );
 396                         if ( $letters === false ) {
 397                                 throw new MWException( "MediaWiki does not support ICU locale " .
 398                                         "\"{$this->locale}\"" );
 399                         }
 400                 }
 401
 402                 // Sort the letters.
 403                 //
 404                 // It's impossible to have the precompiled data file properly sorted,
 405                 // because the sort order changes depending on ICU version. If the
 406                 // array is not properly sorted, the binary search will return random
 407                 // results.
 408                 //
 409                 // We also take this opportunity to remove primary collisions.
 410                 $letterMap = array();
 411                 foreach ( $letters as $letter ) {
 412                         $key = $this->getPrimarySortKey( $letter );
 413                         if ( isset( $letterMap[$key] ) ) {
 414                                 // Primary collision
 415                                 // Keep whichever one sorts first in the main collator
 416                                 if ( $this->mainCollator->compare( $letter, $letterMap[$key] ) < 0 ) {
 417                                         $letterMap[$key] = $letter;
 418                                 }
 419                         } else {
 420                                 $letterMap[$key] = $letter;
 421                         }
 422                 }
 423                 ksort( $letterMap, SORT_STRING );
 424                 // Remove duplicate prefixes. Basically if something has a sortkey
 425                 // which is a prefix of some other sortkey, then it is an
 426                 // expansion and probably should not be considered a section
 427                 // header.
 428                 //
 429                 // For example 'þ' is sometimes sorted as if it is the letters
 430                 // 'th'. Other times it is its own primary element. Another
 431                 // example is '₨'. Sometimes its a currency symbol. Sometimes it
 432                 // is an 'R' followed by an 's'.
 433                 //
 434                 // Additionally an expanded element should always sort directly
 435                 // after its first element due to they way sortkeys work.
 436                 //
 437                 // UCA sortkey elements are of variable length but no collation
 438                 // element should be a prefix of some other element, so I think
 439                 // this is safe. See:
 440                 // * https://ssl.icu-project.org/repos/icu/icuhtml/trunk/design/collation/ICU_collation_design.htm
 441                 // * http://site.icu-project.org/design/collation/uca-weight-allocation
 442                 //
 443                 // Additionally, there is something called primary compression to
 444                 // worry about. Basically, if you have two primary elements that
 445                 // are more than one byte and both start with the same byte then
 446                 // the first byte is dropped on the second primary. Additionally
 447                 // either \x03 or \xFF may be added to mean that the next primary
 448                 // does not start with the first byte of the first primary.
 449                 //
 450                 // This shouldn't matter much, as the first primary is not
 451                 // changed, and that is what we are comparing against.
 452                 //
 453                 // tl;dr: This makes some assumptions about how icu implements
 454                 // collations. It seems incredibly unlikely these assumptions
 455                 // will change, but nonetheless they are assumptions.
 456
 457                 $prev = false;
 458                 $duplicatePrefixes = array();
 459                 foreach ( $letterMap as $key => $value ) {
 460                         // Remove terminator byte. Otherwise the prefix
 461                         // comparison will get hung up on that.
 462                         $trimmedKey = rtrim( $key, "\0" );
 463                         if ( $prev === false || $prev === '' ) {
 464                                 $prev = $trimmedKey;
 465                                 // We don't yet have a collation element
 466                                 // to compare against, so continue.
 467                                 continue;
 468                         }
 469
 470                         // Due to the fact the array is sorted, we only have
 471                         // to compare with the element directly previous
 472                         // to the current element (skipping expansions).
 473                         // An element "X" will always sort directly
 474                         // before "XZ" (Unless we have "XY", but we
 475                         // do not update $prev in that case).
 476                         if ( substr( $trimmedKey, 0, strlen( $prev ) ) === $prev ) {
 477                                 $duplicatePrefixes[] = $key;
 478                                 // If this is an expansion, we don't want to
 479                                 // compare the next element to this element,
 480                                 // but to what is currently $prev
 481                                 continue;
 482                         }
 483                         $prev = $trimmedKey;
 484                 }
 485                 foreach ( $duplicatePrefixes as $badKey ) {
 486                         wfDebug( "Removing '{$letterMap[$badKey]}' from first letters.\n" );
 487                         unset( $letterMap[$badKey] );
 488                         // This code assumes that unsetting does not change sort order.
 489                 }
 490                 $data = array(
 491                         'chars' => array_values( $letterMap ),
 492                         'keys' => array_keys( $letterMap ),
 493                         'version' => self::FIRST_LETTER_VERSION,
 494                 );
 495
 496                 // Reduce memory usage before caching
 497                 unset( $letterMap );
 498
 499                 // Save to cache
 500                 $this->firstLetterData = $data;
 501                 $cache->set( $cacheKey, $data, 86400 * 7 /* 1 week */ );
 502                 return $data;
 503         }
 504
 505         function getLetterByIndex( $index ) {
 506                 if ( $this->firstLetterData === null ) {
 507                         $this->getFirstLetterData();
 508                 }
 509                 return $this->firstLetterData['chars'][$index];
 510         }
 511
 512         function getSortKeyByLetterIndex( $index ) {
 513                 if ( $this->firstLetterData === null ) {
 514                         $this->getFirstLetterData();
 515                 }
 516                 return $this->firstLetterData['keys'][$index];
 517         }
 518
 519         function getFirstLetterCount() {
 520                 if ( $this->firstLetterData === null ) {
 521                         $this->getFirstLetterData();
 522                 }
 523                 return count( $this->firstLetterData['chars'] );
 524         }
 525
 526         /**
 527          * Do a binary search, and return the index of the largest item that sorts
 528          * less than or equal to the target value.
 529          *
 530          * @deprecated since 1.23; use ArrayUtils::findLowerBound() instead
 531          *
 532          * @param array $valueCallback A function to call to get the value with
 533          *     a given array index.
 534          * @param int $valueCount The number of items accessible via $valueCallback,
 535          *     indexed from 0 to $valueCount - 1
 536          * @param array $comparisonCallback A callback to compare two values, returning
 537          *     -1, 0 or 1 in the style of strcmp().
 538          * @param string $target The target value to find.
 539          *
 540          * @return int|bool The item index of the lower bound, or false if the target value
 541          *     sorts before all items.
 542          */
 543         function findLowerBound( $valueCallback, $valueCount, $comparisonCallback, $target ) {
 544                 wfDeprecated( __METHOD__, '1.23' );
 545                 return ArrayUtils::findLowerBound( $valueCallback, $valueCount, $comparisonCallback, $target );
 546         }
 547
 548         static function isCjk( $codepoint ) {
 549                 foreach ( self::$cjkBlocks as $block ) {
 550                         if ( $codepoint >= $block[0] && $codepoint <= $block[1] ) {
 551                                 return true;
 552                         }
 553                 }
 554                 return false;
 555         }
 556
 557         /**
 558          * Return the version of ICU library used by PHP's intl extension,
 559          * or false when the extension is not installed of the version
 560          * can't be determined.
 561          *
 562          * The constant INTL_ICU_VERSION this function refers to isn't really
 563          * documented. It is available since PHP 5.3.7 (see PHP bug 54561).
 564          * This function will return false on older PHPs.
 565          *
 566          * @since 1.21
 567          * @return string|bool
 568          */
 569         static function getICUVersion() {
 570                 return defined( 'INTL_ICU_VERSION' ) ? INTL_ICU_VERSION : false;
 571         }
 572
 573         /**
 574          * Return the version of Unicode appropriate for the version of ICU library
 575          * currently in use, or false when it can't be determined.
 576          *
 577          * @since 1.21
 578          * @return string|bool
 579          */
 580         static function getUnicodeVersionForICU() {
 581                 $icuVersion = IcuCollation::getICUVersion();
 582                 if ( !$icuVersion ) {
 583                         return false;
 584                 }
 585
 586                 $versionPrefix = substr( $icuVersion, 0, 3 );
 587                 // Source: http://site.icu-project.org/download
 588                 $map = array(
 589                         '50.' => '6.2',
 590                         '49.' => '6.1',
 591                         '4.8' => '6.0',
 592                         '4.6' => '6.0',
 593                         '4.4' => '5.2',
 594                         '4.2' => '5.1',
 595                         '4.0' => '5.1',
 596                         '3.8' => '5.0',
 597                         '3.6' => '5.0',
 598                         '3.4' => '4.1',
 599                 );
 600
 601                 if ( isset( $map[$versionPrefix] ) ) {
 602                         return $map[$versionPrefix];
 603                 } else {
 604                         return false;
 605                 }
 606         }
 607 }
 608
 609 /**
 610  * Workaround for the lack of support of Sorani Kurdish / Central Kurdish language ('ckb') in ICU.
 611  *
 612  * Uses the same collation rules as Persian / Farsi ('fa'), but different characters for digits.
 613  */
 614 class CollationCkb extends IcuCollation {
 615         function __construct() {
 616                 // This will set $locale and collators, which affect the actual sorting order
 617                 parent::__construct( 'fa' );
 618                 // Override the 'fa' language set by parent constructor, which affects #getFirstLetterData()
 619                 $this->digitTransformLanguage = Language::factory( 'ckb' );
 620         }
 621 }