includes/collation/IcuCollation.php

   1 <?php
   2 /**
   3  * This program is free software; you can redistribute it and/or modify
   4  * it under the terms of the GNU General Public License as published by
   5  * the Free Software Foundation; either version 2 of the License, or
   6  * (at your option) any later version.
   7  *
   8  * This program is distributed in the hope that it will be useful,
   9  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  10  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  11  * GNU General Public License for more details.
  12  *
  13  * You should have received a copy of the GNU General Public License along
  14  * with this program; if not, write to the Free Software Foundation, Inc.,
  15  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  16  * http://www.gnu.org/copyleft/gpl.html
  17  *
  18  * @file
  19  */
  20
  21 /**
  22  * @since 1.16.3
  23  */
  24 class IcuCollation extends Collation {
  25         const FIRST_LETTER_VERSION = 4;
  26
  27         /** @var Collator */
  28         private $primaryCollator;
  29
  30         /** @var Collator */
  31         private $mainCollator;
  32
  33         /** @var string */
  34         private $locale;
  35
  36         /** @var Language */
  37         protected $digitTransformLanguage;
  38
  39         /** @var bool */
  40         private $useNumericCollation = false;
  41
  42         /** @var array */
  43         private $firstLetterData;
  44
  45         /**
  46          * Unified CJK blocks.
  47          *
  48          * The same definition of a CJK block must be used for both Collation and
  49          * generateCollationData.php. These blocks are omitted from the first
  50          * letter data, as an optimisation measure and because the default UCA table
  51          * is pretty useless for sorting Chinese text anyway. Japanese and Korean
  52          * blocks are not included here, because they are smaller and more useful.
  53          */
  54         private static $cjkBlocks = [
  55                 [ 0x2E80, 0x2EFF ], // CJK Radicals Supplement
  56                 [ 0x2F00, 0x2FDF ], // Kangxi Radicals
  57                 [ 0x2FF0, 0x2FFF ], // Ideographic Description Characters
  58                 [ 0x3000, 0x303F ], // CJK Symbols and Punctuation
  59                 [ 0x31C0, 0x31EF ], // CJK Strokes
  60                 [ 0x3200, 0x32FF ], // Enclosed CJK Letters and Months
  61                 [ 0x3300, 0x33FF ], // CJK Compatibility
  62                 [ 0x3400, 0x4DBF ], // CJK Unified Ideographs Extension A
  63                 [ 0x4E00, 0x9FFF ], // CJK Unified Ideographs
  64                 [ 0xF900, 0xFAFF ], // CJK Compatibility Ideographs
  65                 [ 0xFE30, 0xFE4F ], // CJK Compatibility Forms
  66                 [ 0x20000, 0x2A6DF ], // CJK Unified Ideographs Extension B
  67                 [ 0x2A700, 0x2B73F ], // CJK Unified Ideographs Extension C
  68                 [ 0x2B740, 0x2B81F ], // CJK Unified Ideographs Extension D
  69                 [ 0x2F800, 0x2FA1F ], // CJK Compatibility Ideographs Supplement
  70         ];
  71
  72         /**
  73          * Additional characters (or character groups) to be considered separate
  74          * letters for given languages, or to be removed from the list of such
  75          * letters (denoted by keys starting with '-').
  76          *
  77          * These are additions to (or subtractions from) the data stored in the
  78          * first-letters-root.php data file (which among others includes full basic Latin,
  79          * Cyrillic and Greek alphabets).
  80          *
  81          * "Separate letter" is a letter that would have a separate heading/section
  82          * for it in a dictionary or a phone book in this language. This data isn't
  83          * used for sorting (the ICU library handles that), only for deciding which
  84          * characters (or character groups) to use as headings.
  85          *
  86          * Initially generated based on the primary level of Unicode collation
  87          * tailorings available at http://developer.mimer.com/charts/tailorings.htm ,
  88          * later modified.
  89          *
  90          * Empty arrays are intended; this signifies that the data for the language is
  91          * available and that there are, in fact, no additional letters to consider.
  92          */
  93         private static $tailoringFirstLetters = [
  94                 'af' => [],
  95                 'am' => [],
  96                 'ar' => [],
  97                 'as' => [ "\u{0982}", "\u{0981}", "\u{0983}", "\u{09CE}", "ক্ষ " ],
  98                 'ast' => [ "Ch", "Ll", "Ñ" ], // not in libicu
  99                 'az' => [ "Ç", "Ə", "Ğ", "İ", "Ö", "Ş", "Ü" ],
 100                 'be' => [ "Ё" ],
 101                 'be-tarask' => [ "Ё" ],
 102                 'bg' => [],
 103                 'bn' => [ 'ং', 'ঃ', 'ঁ' ],
 104                 'bn@collation=traditional' => [
 105                         'ং', 'ঃ', 'ঁ', 'ক্', 'খ্', 'গ্', 'ঘ্', 'ঙ্', 'চ্', 'ছ্', 'জ্', 'ঝ্',
 106                         'ঞ্', 'ট্', 'ঠ্', 'ড্', 'ঢ্', 'ণ্', 'ৎ', 'থ্', 'দ্', 'ধ্', 'ন্', 'প্',
 107                         'ফ্', 'ব্', 'ভ্', 'ম্', 'য্', 'র্', 'ৰ্', 'ল্', 'ৱ্', 'শ্', 'ষ্', 'স্', 'হ্'
 108                 ],
 109                 'bo' => [],
 110                 'br' => [ "Ch", "C'h" ],
 111                 'bs' => [ "Č", "Ć", "Dž", "Đ", "Lj", "Nj", "Š", "Ž" ],
 112                 'bs-Cyrl' => [],
 113                 'ca' => [],
 114                 'chr' => [],
 115                 'co' => [], // not in libicu
 116                 'cs' => [ "Č", "Ch", "Ř", "Š", "Ž" ],
 117                 'cy' => [ "Ch", "Dd", "Ff", "Ng", "Ll", "Ph", "Rh", "Th" ],
 118                 'da' => [ "Æ", "Ø", "Å" ],
 119                 'de' => [],
 120                 'de-AT@collation=phonebook' => [ 'ä', 'ö', 'ü', 'ß' ],
 121                 'dsb' => [ "Č", "Ć", "Dź", "Ě", "Ch", "Ł", "Ń", "Ŕ", "Š", "Ś", "Ž", "Ź" ],
 122                 'ee' => [ "Dz", "Ɖ", "Ɛ", "Ƒ", "Gb", "Ɣ", "Kp", "Ny", "Ŋ", "Ɔ", "Ts", "Ʋ" ],
 123                 'el' => [],
 124                 'en' => [],
 125                 'eo' => [ "Ĉ", "Ĝ", "Ĥ", "Ĵ", "Ŝ", "Ŭ" ],
 126                 'es' => [ "Ñ" ],
 127                 'et' => [ "Š", "Ž", "Õ", "Ä", "Ö", "Ü" ],
 128                 'eu' => [ "Ñ" ], // not in libicu
 129                 'fa' => [
 130                         // RTL, let's put each letter on a new line
 131                         "آ",
 132                         "ء",
 133                         "ه",
 134                         "ا",
 135                         "و"
 136                 ],
 137                 'fi' => [ "Å", "Ä", "Ö" ],
 138                 'fil' => [ "Ñ", "Ng" ],
 139                 'fo' => [ "Á", "Ð", "Í", "Ó", "Ú", "Ý", "Æ", "Ø", "Å" ],
 140                 'fr' => [],
 141                 'fr-CA' => [], // fr-CA sorts accents slightly different from fr.
 142                 'fur' => [ "À", "Á", "Â", "È", "Ì", "Ò", "Ù" ], // not in libicu
 143                 'fy' => [], // not in libicu
 144                 'ga' => [],
 145                 'gd' => [], // not in libicu
 146                 'gl' => [ "Ch", "Ll", "Ñ" ],
 147                 'gu' => [ "\u{0A82}", "\u{0A83}", "\u{0A81}", "\u{0AB3}" ],
 148                 'ha' => [ 'Ɓ', 'Ɗ', 'Ƙ', 'Sh', 'Ts', 'Ƴ' ],
 149                 'haw' => [ 'ʻ' ],
 150                 'he' => [],
 151                 'hi' => [ "\u{0902}", "\u{0903}" ],
 152                 'hr' => [ "Č", "Ć", "Dž", "Đ", "Lj", "Nj", "Š", "Ž" ],
 153                 'hsb' => [ "Č", "Dź", "Ě", "Ch", "Ł", "Ń", "Ř", "Š", "Ć", "Ž" ],
 154                 'hu' => [ "Cs", "Dz", "Dzs", "Gy", "Ly", "Ny", "Ö", "Sz", "Ty", "Ü", "Zs" ],
 155                 'hy' => [ "և" ],
 156                 'id' => [],
 157                 'ig' => [ "Ch", "Gb", "Gh", "Gw", "Ị", "Kp", "Kw", "Ṅ", "Nw", "Ny", "Ọ", "Sh", "Ụ" ],
 158                 'is' => [ "Á", "Ð", "É", "Í", "Ó", "Ú", "Ý", "Þ", "Æ", "Ö", "Å" ],
 159                 'it' => [],
 160                 'ka' => [],
 161                 'kk' => [ "Ү", "І" ],
 162                 'kl' => [ "Æ", "Ø", "Å" ],
 163                 'km' => [
 164                         "រ", "ឫ", "ឬ", "ល", "ឭ", "ឮ", "\u{17BB}\u{17C6}",
 165                         "\u{17C6}", "\u{17B6}\u{17C6}", "\u{17C7}",
 166                         "\u{17B7}\u{17C7}", "\u{17BB}\u{17C7}",
 167                         "\u{17C1}\u{17C7}", "\u{17C4}\u{17C7}",
 168                 ],
 169                 'kn' => [ "\u{0C81}", "\u{0C83}", "\u{0CF1}", "\u{0CF2}" ],
 170                 'kok' => [ "\u{0902}", "\u{0903}", "ळ", "क्ष" ],
 171                 'ku' => [ "Ç", "Ê", "Î", "Ş", "Û" ], // not in libicu
 172                 'ky' => [ "Ё" ],
 173                 'la' => [], // not in libicu
 174                 'lb' => [],
 175                 'lkt' => [ 'Č', 'Ǧ', 'Ȟ', 'Š', 'Ž' ],
 176                 'ln' => [ 'Ɛ' ],
 177                 'lo' => [],
 178                 'lt' => [ "Č", "Š", "Ž" ],
 179                 'lv' => [ "Č", "Ģ", "Ķ", "Ļ", "Ņ", "Š", "Ž" ],
 180                 'mk' => [ "Ѓ", "Ќ" ],
 181                 'ml' => [],
 182                 'mn' => [],
 183                 'mo' => [ "Ă", "Â", "Î", "Ș", "Ț" ], // not in libicu
 184                 'mr' => [ "\u{0902}", "\u{0903}", "ळ", "क्ष", "ज्ञ" ],
 185                 'ms' => [],
 186                 'mt' => [ "Ċ", "Ġ", "Għ", "Ħ", "Ż" ],
 187                 'nb' => [ "Æ", "Ø", "Å" ],
 188                 'ne' => [],
 189                 'nl' => [],
 190                 'nn' => [ "Æ", "Ø", "Å" ],
 191                 'no' => [ "Æ", "Ø", "Å" ], // not in libicu. You should probably use nb or nn instead.
 192                 'oc' => [], // not in libicu
 193                 'om' => [ 'Ch', 'Dh', 'Kh', 'Ny', 'Ph', 'Sh' ],
 194                 'or' => [ "\u{0B01}", "\u{0B02}", "\u{0B03}", "କ୍ଷ" ],
 195                 'pa' => [ "\u{0A4D}" ],
 196                 'pl' => [ "Ą", "Ć", "Ę", "Ł", "Ń", "Ó", "Ś", "Ź", "Ż" ],
 197                 'pt' => [],
 198                 'rm' => [], // not in libicu
 199                 'ro' => [ "Ă", "Â", "Î", "Ș", "Ț" ],
 200                 'ru' => [],
 201                 'rup' => [ "Ă", "Â", "Î", "Ľ", "Ń", "Ș", "Ț" ], // not in libicu
 202                 'sco' => [],
 203                 'se' => [
 204                         'Á', 'Č', 'Ʒ', 'Ǯ', 'Đ', 'Ǧ', 'Ǥ', 'Ǩ', 'Ŋ',
 205                         'Š', 'Ŧ', 'Ž', 'Ø', 'Æ', 'Ȧ', 'Ä', 'Ö'
 206                 ],
 207                 'si' => [ "\u{0D82}", "\u{0D83}", "\u{0DA4}" ],
 208                 'sk' => [ "Ä", "Č", "Ch", "Ô", "Š", "Ž" ],
 209                 'sl' => [ "Č", "Š", "Ž" ],
 210                 'smn' => [ "Á", "Č", "Đ", "Ŋ", "Š", "Ŧ", "Ž", "Æ", "Ø", "Å", "Ä", "Ö" ],
 211                 'sq' => [ "Ç", "Dh", "Ë", "Gj", "Ll", "Nj", "Rr", "Sh", "Th", "Xh", "Zh" ],
 212                 'sr' => [],
 213                 'sr-Latn' => [ "Č", "Ć", "Dž", "Đ", "Lj", "Nj", "Š", "Ž" ],
 214                 'sv' => [ "Å", "Ä", "Ö" ],
 215                 'sv@collation=standard' => [ "Å", "Ä", "Ö" ],
 216                 'sw' => [],
 217                 'ta' => [
 218                         "\u{0B82}", "ஃ", "க்ஷ", "க்", "ங்", "ச்", "ஞ்", "ட்", "ண்", "த்", "ந்",
 219                         "ப்", "ம்", "ய்", "ர்", "ல்", "வ்", "ழ்", "ள்", "ற்", "ன்", "ஜ்", "ஶ்", "ஷ்",
 220                         "ஸ்", "ஹ்", "க்ஷ்"
 221                 ],
 222                 'te' => [ "\u{0C01}", "\u{0C02}", "\u{0C03}" ],
 223                 'th' => [ "ฯ", "\u{0E46}", "\u{0E4D}", "\u{0E3A}" ],
 224                 'tk' => [ "Ç", "Ä", "Ž", "Ň", "Ö", "Ş", "Ü", "Ý" ],
 225                 'tl' => [ "Ñ", "Ng" ], // not in libicu
 226                 'to' => [ "Ng", "ʻ" ],
 227                 'tr' => [ "Ç", "Ğ", "İ", "Ö", "Ş", "Ü" ],
 228                 '-tr' => [ "ı" ],
 229                 'tt' => [ "Ә", "Ө", "Ү", "Җ", "Ң", "Һ" ], // not in libicu
 230                 'uk' => [ "Ґ", "Ь" ],
 231                 'uz' => [ "Ch", "G'", "Ng", "O'", "Sh" ], // not in libicu
 232                 'vi' => [ "Ă", "Â", "Đ", "Ê", "Ô", "Ơ", "Ư" ],
 233                 'vo' => [ "Ä", "Ö", "Ü" ],
 234                 'yi' => [
 235                         "\u{05D1}\u{05BF}", "\u{05DB}\u{05BC}", "\u{05E4}\u{05BC}",
 236                         "\u{05E9}\u{05C2}", "\u{05EA}\u{05BC}"
 237                 ],
 238                 'yo' => [ "Ẹ", "Gb", "Ọ", "Ṣ" ],
 239                 'zu' => [],
 240         ];
 241
 242         /**
 243          * @since 1.16.3
 244          */
 245         const RECORD_LENGTH = 14;
 246
 247         public function __construct( $locale ) {
 248                 if ( !extension_loaded( 'intl' ) ) {
 249                         throw new MWException( 'An ICU collation was requested, ' .
 250                                 'but the intl extension is not available.' );
 251                 }
 252
 253                 $this->locale = $locale;
 254                 // Drop everything after the '@' in locale's name
 255                 $localeParts = explode( '@', $locale );
 256                 $this->digitTransformLanguage = Language::factory( $locale === 'root' ? 'en' : $localeParts[0] );
 257
 258                 $this->mainCollator = Collator::create( $locale );
 259                 if ( !$this->mainCollator ) {
 260                         throw new MWException( "Invalid ICU locale specified for collation: $locale" );
 261                 }
 262
 263                 $this->primaryCollator = Collator::create( $locale );
 264                 $this->primaryCollator->setStrength( Collator::PRIMARY );
 265
 266                 // If the special suffix for numeric collation is present, turn on numeric collation.
 267                 if ( substr( $locale, -5, 5 ) === '-u-kn' ) {
 268                         $this->useNumericCollation = true;
 269                         // Strip off the special suffix so it doesn't trip up fetchFirstLetterData().
 270                         $this->locale = substr( $this->locale, 0, -5 );
 271                         $this->mainCollator->setAttribute( Collator::NUMERIC_COLLATION, Collator::ON );
 272                         $this->primaryCollator->setAttribute( Collator::NUMERIC_COLLATION, Collator::ON );
 273                 }
 274         }
 275
 276         public function getSortKey( $string ) {
 277                 return $this->mainCollator->getSortKey( $string );
 278         }
 279
 280         public function getPrimarySortKey( $string ) {
 281                 return $this->primaryCollator->getSortKey( $string );
 282         }
 283
 284         public function getFirstLetter( $string ) {
 285                 $string = strval( $string );
 286                 if ( $string === '' ) {
 287                         return '';
 288                 }
 289
 290                 $firstChar = mb_substr( $string, 0, 1, 'UTF-8' );
 291
 292                 // If the first character is a CJK character, just return that character.
 293                 if ( ord( $firstChar ) > 0x7f && self::isCjk( UtfNormal\Utils::utf8ToCodepoint( $firstChar ) ) ) {
 294                         return $firstChar;
 295                 }
 296
 297                 $sortKey = $this->getPrimarySortKey( $string );
 298
 299                 // Do a binary search to find the correct letter to sort under
 300                 $min = ArrayUtils::findLowerBound(
 301                         [ $this, 'getSortKeyByLetterIndex' ],
 302                         $this->getFirstLetterCount(),
 303                         'strcmp',
 304                         $sortKey );
 305
 306                 if ( $min === false ) {
 307                         // Before the first letter
 308                         return '';
 309                 }
 310
 311                 $sortLetter = $this->getLetterByIndex( $min );
 312
 313                 if ( $this->useNumericCollation ) {
 314                         // If the sort letter is a number, return '0–9' (or localized equivalent).
 315                         // ASCII value of 0 is 48. ASCII value of 9 is 57.
 316                         // Note that this also applies to non-Arabic numerals since they are
 317                         // mapped to Arabic numeral sort letters. For example, ২ sorts as 2.
 318                         if ( ord( $sortLetter ) >= 48 && ord( $sortLetter ) <= 57 ) {
 319                                 $sortLetter = wfMessage( 'category-header-numerals' )->numParams( 0, 9 )->text();
 320                         }
 321                 }
 322                 return $sortLetter;
 323         }
 324
 325         /**
 326          * @since 1.16.3
 327          * @return array
 328          */
 329         public function getFirstLetterData() {
 330                 if ( $this->firstLetterData === null ) {
 331                         $cache = ObjectCache::getLocalServerInstance( CACHE_ANYTHING );
 332                         $cacheKey = $cache->makeKey(
 333                                 'first-letters',
 334                                 static::class,
 335                                 $this->locale,
 336                                 $this->digitTransformLanguage->getCode(),
 337                                 INTL_ICU_VERSION,
 338                                 self::FIRST_LETTER_VERSION
 339                         );
 340                         $this->firstLetterData = $cache->getWithSetCallback( $cacheKey, $cache::TTL_WEEK, function () {
 341                                 return $this->fetchFirstLetterData();
 342                         } );
 343                 }
 344                 return $this->firstLetterData;
 345         }
 346
 347         /**
 348          * @return array
 349          * @throws MWException
 350          */
 351         private function fetchFirstLetterData() {
 352                 global $IP;
 353                 // Generate data from serialized data file
 354                 if ( isset( self::$tailoringFirstLetters[$this->locale] ) ) {
 355                         $letters = require "$IP/includes/collation/data/first-letters-root.php";
 356                         // Append additional characters
 357                         $letters = array_merge( $letters, self::$tailoringFirstLetters[$this->locale] );
 358                         // Remove unnecessary ones, if any
 359                         if ( isset( self::$tailoringFirstLetters['-' . $this->locale] ) ) {
 360                                 $letters = array_diff( $letters, self::$tailoringFirstLetters['-' . $this->locale] );
 361                         }
 362                         // Apply digit transforms
 363                         $digits = [ '0', '1', '2', '3', '4', '5', '6', '7', '8', '9' ];
 364                         $letters = array_diff( $letters, $digits );
 365                         foreach ( $digits as $digit ) {
 366                                 $letters[] = $this->digitTransformLanguage->formatNum( $digit, true );
 367                         }
 368                 } elseif ( $this->locale === 'root' ) {
 369                         $letters = require "$IP/includes/collation/data/first-letters-root.php";
 370                 } else {
 371                         // FIXME: Is this still used?
 372                         $letters = wfGetPrecompiledData( "first-letters-{$this->locale}.ser" );
 373                         if ( $letters === false ) {
 374                                 throw new MWException( "MediaWiki does not support ICU locale " .
 375                                         "\"{$this->locale}\"" );
 376                         }
 377                 }
 378
 379                 /* Sort the letters.
 380                  *
 381                  * It's impossible to have the precompiled data file properly sorted,
 382                  * because the sort order changes depending on ICU version. If the
 383                  * array is not properly sorted, the binary search will return random
 384                  * results.
 385                  *
 386                  * We also take this opportunity to remove primary collisions.
 387                  */
 388                 $letterMap = [];
 389                 foreach ( $letters as $letter ) {
 390                         $key = $this->getPrimarySortKey( $letter );
 391                         if ( isset( $letterMap[$key] ) ) {
 392                                 // Primary collision (two characters with the same sort position).
 393                                 // Keep whichever one sorts first in the main collator.
 394                                 $comp = $this->mainCollator->compare( $letter, $letterMap[$key] );
 395                                 wfDebug( "Primary collision '$letter' '{$letterMap[$key]}' (comparison: $comp)\n" );
 396                                 // If that also has a collision, use codepoint as a tiebreaker.
 397                                 if ( $comp === 0 ) {
 398                                         $comp = UtfNormal\Utils::utf8ToCodepoint( $letter ) <=>
 399                                                 UtfNormal\Utils::utf8ToCodepoint( $letterMap[$key] );
 400                                 }
 401                                 if ( $comp < 0 ) {
 402                                         $letterMap[$key] = $letter;
 403                                 }
 404                         } else {
 405                                 $letterMap[$key] = $letter;
 406                         }
 407                 }
 408                 ksort( $letterMap, SORT_STRING );
 409
 410                 /* Remove duplicate prefixes. Basically if something has a sortkey
 411                  * which is a prefix of some other sortkey, then it is an
 412                  * expansion and probably should not be considered a section
 413                  * header.
 414                  *
 415                  * For example 'þ' is sometimes sorted as if it is the letters
 416                  * 'th'. Other times it is its own primary element. Another
 417                  * example is '₨'. Sometimes its a currency symbol. Sometimes it
 418                  * is an 'R' followed by an 's'.
 419                  *
 420                  * Additionally an expanded element should always sort directly
 421                  * after its first element due to they way sortkeys work.
 422                  *
 423                  * UCA sortkey elements are of variable length but no collation
 424                  * element should be a prefix of some other element, so I think
 425                  * this is safe. See:
 426                  * - https://ssl.icu-project.org/repos/icu/icuhtml/trunk/design/collation/ICU_collation_design.htm
 427                  * - http://site.icu-project.org/design/collation/uca-weight-allocation
 428                  *
 429                  * Additionally, there is something called primary compression to
 430                  * worry about. Basically, if you have two primary elements that
 431                  * are more than one byte and both start with the same byte then
 432                  * the first byte is dropped on the second primary. Additionally
 433                  * either \x03 or \xFF may be added to mean that the next primary
 434                  * does not start with the first byte of the first primary.
 435                  *
 436                  * This shouldn't matter much, as the first primary is not
 437                  * changed, and that is what we are comparing against.
 438                  *
 439                  * tl;dr: This makes some assumptions about how icu implements
 440                  * collations. It seems incredibly unlikely these assumptions
 441                  * will change, but nonetheless they are assumptions.
 442                  */
 443
 444                 $prev = false;
 445                 $duplicatePrefixes = [];
 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.\n" );
 474                         unset( $letterMap[$badKey] );
 475                         // This code assumes that unsetting does not change sort order.
 476                 }
 477                 $data = [
 478                         'chars' => array_values( $letterMap ),
 479                         'keys' => array_keys( $letterMap ),
 480                 ];
 481
 482                 // Reduce memory usage before caching
 483                 unset( $letterMap );
 484
 485                 return $data;
 486         }
 487
 488         /**
 489          * @param string $index
 490          * @return string
 491          * @since 1.16.3
 492          */
 493         public function getLetterByIndex( $index ) {
 494                 return $this->getFirstLetterData()['chars'][$index];
 495         }
 496
 497         /**
 498          * @param string $index
 499          * @return string
 500          * @since 1.16.3
 501          */
 502         public function getSortKeyByLetterIndex( $index ) {
 503                 return $this->getFirstLetterData()['keys'][$index];
 504         }
 505
 506         /**
 507          * @return string
 508          * @since 1.16.3
 509          */
 510         public function getFirstLetterCount() {
 511                 return count( $this->getFirstLetterData()['chars'] );
 512         }
 513
 514         /**
 515          * Test if a code point is a CJK (Chinese, Japanese, Korean) character
 516          * @param int $codepoint
 517          * @return bool
 518          * @since 1.16.3
 519          */
 520         public static function isCjk( $codepoint ) {
 521                 foreach ( self::$cjkBlocks as $block ) {
 522                         if ( $codepoint >= $block[0] && $codepoint <= $block[1] ) {
 523                                 return true;
 524                         }
 525                 }
 526                 return false;
 527         }
 528
 529         /**
 530          * Return the version of Unicode appropriate for the version of ICU library
 531          * currently in use, or false when it can't be determined.
 532          *
 533          * @since 1.21
 534          * @return string|bool
 535          */
 536         static function getUnicodeVersionForICU() {
 537                 $icuVersion = INTL_ICU_VERSION;
 538                 if ( !$icuVersion ) {
 539                         return false;
 540                 }
 541
 542                 $versionPrefix = substr( $icuVersion, 0, 3 );
 543                 // Source: http://site.icu-project.org/download
 544                 $map = [
 545                         '63.' => '11.0',
 546                         '62.' => '11.0',
 547                         '61.' => '10.0',
 548                         '60.' => '10.0',
 549                         '59.' => '9.0',
 550                         '58.' => '9.0',
 551                         '57.' => '8.0',
 552                         '56.' => '8.0',
 553                         '55.' => '7.0',
 554                         '54.' => '7.0',
 555                         '53.' => '6.3',
 556                         '52.' => '6.3',
 557                         '51.' => '6.2',
 558                         '50.' => '6.2',
 559                         '49.' => '6.1',
 560                         '4.8' => '6.0',
 561                         '4.6' => '6.0',
 562                         '4.4' => '5.2',
 563                         '4.2' => '5.1',
 564                         '4.0' => '5.1',
 565                         '3.8' => '5.0',
 566                         '3.6' => '5.0',
 567                         '3.4' => '4.1',
 568                 ];
 569
 570                 return $map[$versionPrefix] ?? false;
 571         }
 572 }