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