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.
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.
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
24 class IcuCollation
extends Collation
{
25 const FIRST_LETTER_VERSION
= 2;
28 private $primaryCollator;
31 private $mainCollator;
37 protected $digitTransformLanguage;
40 private $firstLetterData;
45 * The same definition of a CJK block must be used for both Collation and
46 * generateCollationData.php. These blocks are omitted from the first
47 * letter data, as an optimisation measure and because the default UCA table
48 * is pretty useless for sorting Chinese text anyway. Japanese and Korean
49 * blocks are not included here, because they are smaller and more useful.
51 private static $cjkBlocks = [
52 [ 0x2E80, 0x2EFF ], // CJK Radicals Supplement
53 [ 0x2F00, 0x2FDF ], // Kangxi Radicals
54 [ 0x2FF0, 0x2FFF ], // Ideographic Description Characters
55 [ 0x3000, 0x303F ], // CJK Symbols and Punctuation
56 [ 0x31C0, 0x31EF ], // CJK Strokes
57 [ 0x3200, 0x32FF ], // Enclosed CJK Letters and Months
58 [ 0x3300, 0x33FF ], // CJK Compatibility
59 [ 0x3400, 0x4DBF ], // CJK Unified Ideographs Extension A
60 [ 0x4E00, 0x9FFF ], // CJK Unified Ideographs
61 [ 0xF900, 0xFAFF ], // CJK Compatibility Ideographs
62 [ 0xFE30, 0xFE4F ], // CJK Compatibility Forms
63 [ 0x20000, 0x2A6DF ], // CJK Unified Ideographs Extension B
64 [ 0x2A700, 0x2B73F ], // CJK Unified Ideographs Extension C
65 [ 0x2B740, 0x2B81F ], // CJK Unified Ideographs Extension D
66 [ 0x2F800, 0x2FA1F ], // CJK Compatibility Ideographs Supplement
70 * Additional characters (or character groups) to be considered separate
71 * letters for given languages, or to be removed from the list of such
72 * letters (denoted by keys starting with '-').
74 * These are additions to (or subtractions from) the data stored in the
75 * first-letters-root.ser file (which among others includes full basic latin,
76 * cyrillic and greek alphabets).
78 * "Separate letter" is a letter that would have a separate heading/section
79 * for it in a dictionary or a phone book in this language. This data isn't
80 * used for sorting (the ICU library handles that), only for deciding which
81 * characters (or character groups) to use as headings.
83 * Initially generated based on the primary level of Unicode collation
84 * tailorings available at http://developer.mimer.com/charts/tailorings.htm ,
87 * Empty arrays are intended; this signifies that the data for the language is
88 * available and that there are, in fact, no additional letters to consider.
90 private static $tailoringFirstLetters = [
91 // Verified by native speakers
93 'be-tarask' => [ "Ё" ],
94 'cy' => [ "Ch", "Dd", "Ff", "Ng", "Ll", "Ph", "Rh", "Th" ],
96 'fa' => [ "آ", "ء", "ه" ],
97 'fi' => [ "Å", "Ä", "Ö" ],
99 'hu' => [ "Cs", "Dz", "Dzs", "Gy", "Ly", "Ny", "Ö", "Sz", "Ty", "Ü", "Zs" ],
100 'is' => [ "Á", "Ð", "É", "Í", "Ó", "Ú", "Ý", "Þ", "Æ", "Ö", "Å" ],
102 'lv' => [ "Č", "Ģ", "Ķ", "Ļ", "Ņ", "Š", "Ž" ],
103 'pl' => [ "Ą", "Ć", "Ę", "Ł", "Ń", "Ó", "Ś", "Ź", "Ż" ],
106 'sv' => [ "Å", "Ä", "Ö" ],
107 'sv@collation=standard' => [ "Å", "Ä", "Ö" ],
108 'uk' => [ "Ґ", "Ь" ],
109 'vi' => [ "Ă", "Â", "Đ", "Ê", "Ô", "Ơ", "Ư" ],
110 // Not verified, but likely correct
112 'ast' => [ "Ch", "Ll", "Ñ" ],
113 'az' => [ "Ç", "Ə", "Ğ", "İ", "Ö", "Ş", "Ü" ],
115 'br' => [ "Ch", "C'h" ],
116 'bs' => [ "Č", "Ć", "Dž", "Đ", "Lj", "Nj", "Š", "Ž" ],
119 'cs' => [ "Č", "Ch", "Ř", "Š", "Ž" ],
120 'da' => [ "Æ", "Ø", "Å" ],
122 'dsb' => [ "Č", "Ć", "Dź", "Ě", "Ch", "Ł", "Ń", "Ŕ", "Š", "Ś", "Ž", "Ź" ],
124 'eo' => [ "Ĉ", "Ĝ", "Ĥ", "Ĵ", "Ŝ", "Ŭ" ],
126 'et' => [ "Š", "Ž", "Õ", "Ä", "Ö", "Ü", "W" ], // added W for CollationEt (xx-uca-et)
128 'fo' => [ "Á", "Ð", "Í", "Ó", "Ú", "Ý", "Æ", "Ø", "Å" ],
129 'fur' => [ "À", "Á", "Â", "È", "Ì", "Ò", "Ù" ],
133 'gl' => [ "Ch", "Ll", "Ñ" ],
134 'hr' => [ "Č", "Ć", "Dž", "Đ", "Lj", "Nj", "Š", "Ž" ],
135 'hsb' => [ "Č", "Dź", "Ě", "Ch", "Ł", "Ń", "Ř", "Š", "Ć", "Ž" ],
136 'kk' => [ "Ү", "І" ],
137 'kl' => [ "Æ", "Ø", "Å" ],
138 'ku' => [ "Ç", "Ê", "Î", "Ş", "Û" ],
142 'lt' => [ "Č", "Š", "Ž" ],
144 'mo' => [ "Ă", "Â", "Î", "Ş", "Ţ" ],
145 'mt' => [ "Ċ", "Ġ", "Għ", "Ħ", "Ż" ],
147 'no' => [ "Æ", "Ø", "Å" ],
150 'ro' => [ "Ă", "Â", "Î", "Ş", "Ţ" ],
151 'rup' => [ "Ă", "Â", "Î", "Ľ", "Ń", "Ş", "Ţ" ],
153 'sk' => [ "Ä", "Č", "Ch", "Ô", "Š", "Ž" ],
154 'sl' => [ "Č", "Š", "Ž" ],
155 'smn' => [ "Á", "Č", "Đ", "Ŋ", "Š", "Ŧ", "Ž", "Æ", "Ø", "Å", "Ä", "Ö" ],
156 'sq' => [ "Ç", "Dh", "Ë", "Gj", "Ll", "Nj", "Rr", "Sh", "Th", "Xh", "Zh" ],
158 'tk' => [ "Ç", "Ä", "Ž", "Ň", "Ö", "Ş", "Ü", "Ý" ],
159 'tl' => [ "Ñ", "Ng" ],
160 'tr' => [ "Ç", "Ğ", "İ", "Ö", "Ş", "Ü" ],
161 'tt' => [ "Ә", "Ө", "Ү", "Җ", "Ң", "Һ" ],
162 'uz' => [ "Ch", "G'", "Ng", "O'", "Sh" ],
168 const RECORD_LENGTH
= 14;
170 public function __construct( $locale ) {
171 if ( !extension_loaded( 'intl' ) ) {
172 throw new MWException( 'An ICU collation was requested, ' .
173 'but the intl extension is not available.' );
176 $this->locale
= $locale;
177 // Drop everything after the '@' in locale's name
178 $localeParts = explode( '@', $locale );
179 $this->digitTransformLanguage
= Language
::factory( $locale === 'root' ?
'en' : $localeParts[0] );
181 $this->mainCollator
= Collator
::create( $locale );
182 if ( !$this->mainCollator
) {
183 throw new MWException( "Invalid ICU locale specified for collation: $locale" );
186 $this->primaryCollator
= Collator
::create( $locale );
187 $this->primaryCollator
->setStrength( Collator
::PRIMARY
);
190 public function getSortKey( $string ) {
191 return $this->mainCollator
->getSortKey( $string );
194 public function getPrimarySortKey( $string ) {
195 return $this->primaryCollator
->getSortKey( $string );
198 public function getFirstLetter( $string ) {
199 $string = strval( $string );
200 if ( $string === '' ) {
205 $firstChar = mb_substr( $string, 0, 1, 'UTF-8' );
206 if ( ord( $firstChar ) > 0x7f && self
::isCjk( UtfNormal\Utils
::utf8ToCodepoint( $firstChar ) ) ) {
210 $sortKey = $this->getPrimarySortKey( $string );
212 // Do a binary search to find the correct letter to sort under
213 $min = ArrayUtils
::findLowerBound(
214 [ $this, 'getSortKeyByLetterIndex' ],
215 $this->getFirstLetterCount(),
219 if ( $min === false ) {
220 // Before the first letter
223 return $this->getLetterByIndex( $min );
230 public function getFirstLetterData() {
231 if ( $this->firstLetterData
=== null ) {
232 $cache = ObjectCache
::getLocalServerInstance( CACHE_ANYTHING
);
233 $cacheKey = $cache->makeKey(
236 $this->digitTransformLanguage
->getCode(),
237 self
::getICUVersion(),
238 self
::FIRST_LETTER_VERSION
240 $this->firstLetterData
= $cache->getWithSetCallback( $cacheKey, $cache::TTL_WEEK
, function () {
241 return $this->fetchFirstLetterData();
244 return $this->firstLetterData
;
249 * @throws MWException
251 private function fetchFirstLetterData() {
252 // Generate data from serialized data file
253 if ( isset( self
::$tailoringFirstLetters[$this->locale
] ) ) {
254 $letters = wfGetPrecompiledData( 'first-letters-root.ser' );
255 // Append additional characters
256 $letters = array_merge( $letters, self
::$tailoringFirstLetters[$this->locale
] );
257 // Remove unnecessary ones, if any
258 if ( isset( self
::$tailoringFirstLetters['-' . $this->locale
] ) ) {
259 $letters = array_diff( $letters, self
::$tailoringFirstLetters['-' . $this->locale
] );
261 // Apply digit transforms
262 $digits = [ '0', '1', '2', '3', '4', '5', '6', '7', '8', '9' ];
263 $letters = array_diff( $letters, $digits );
264 foreach ( $digits as $digit ) {
265 $letters[] = $this->digitTransformLanguage
->formatNum( $digit, true );
268 $letters = wfGetPrecompiledData( "first-letters-{$this->locale}.ser" );
269 if ( $letters === false ) {
270 throw new MWException( "MediaWiki does not support ICU locale " .
271 "\"{$this->locale}\"" );
277 * It's impossible to have the precompiled data file properly sorted,
278 * because the sort order changes depending on ICU version. If the
279 * array is not properly sorted, the binary search will return random
282 * We also take this opportunity to remove primary collisions.
285 foreach ( $letters as $letter ) {
286 $key = $this->getPrimarySortKey( $letter );
287 if ( isset( $letterMap[$key] ) ) {
289 // Keep whichever one sorts first in the main collator
290 if ( $this->mainCollator
->compare( $letter, $letterMap[$key] ) < 0 ) {
291 $letterMap[$key] = $letter;
294 $letterMap[$key] = $letter;
297 ksort( $letterMap, SORT_STRING
);
299 /* Remove duplicate prefixes. Basically if something has a sortkey
300 * which is a prefix of some other sortkey, then it is an
301 * expansion and probably should not be considered a section
304 * For example 'þ' is sometimes sorted as if it is the letters
305 * 'th'. Other times it is its own primary element. Another
306 * example is '₨'. Sometimes its a currency symbol. Sometimes it
307 * is an 'R' followed by an 's'.
309 * Additionally an expanded element should always sort directly
310 * after its first element due to they way sortkeys work.
312 * UCA sortkey elements are of variable length but no collation
313 * element should be a prefix of some other element, so I think
315 * - https://ssl.icu-project.org/repos/icu/icuhtml/trunk/design/collation/ICU_collation_design.htm
316 * - http://site.icu-project.org/design/collation/uca-weight-allocation
318 * Additionally, there is something called primary compression to
319 * worry about. Basically, if you have two primary elements that
320 * are more than one byte and both start with the same byte then
321 * the first byte is dropped on the second primary. Additionally
322 * either \x03 or \xFF may be added to mean that the next primary
323 * does not start with the first byte of the first primary.
325 * This shouldn't matter much, as the first primary is not
326 * changed, and that is what we are comparing against.
328 * tl;dr: This makes some assumptions about how icu implements
329 * collations. It seems incredibly unlikely these assumptions
330 * will change, but nonetheless they are assumptions.
334 $duplicatePrefixes = [];
335 foreach ( $letterMap as $key => $value ) {
336 // Remove terminator byte. Otherwise the prefix
337 // comparison will get hung up on that.
338 $trimmedKey = rtrim( $key, "\0" );
339 if ( $prev === false ||
$prev === '' ) {
341 // We don't yet have a collation element
342 // to compare against, so continue.
346 // Due to the fact the array is sorted, we only have
347 // to compare with the element directly previous
348 // to the current element (skipping expansions).
349 // An element "X" will always sort directly
350 // before "XZ" (Unless we have "XY", but we
351 // do not update $prev in that case).
352 if ( substr( $trimmedKey, 0, strlen( $prev ) ) === $prev ) {
353 $duplicatePrefixes[] = $key;
354 // If this is an expansion, we don't want to
355 // compare the next element to this element,
356 // but to what is currently $prev
361 foreach ( $duplicatePrefixes as $badKey ) {
362 wfDebug( "Removing '{$letterMap[$badKey]}' from first letters.\n" );
363 unset( $letterMap[$badKey] );
364 // This code assumes that unsetting does not change sort order.
367 'chars' => array_values( $letterMap ),
368 'keys' => array_keys( $letterMap ),
371 // Reduce memory usage before caching
380 public function getLetterByIndex( $index ) {
381 return $this->getFirstLetterData()['chars'][$index];
387 public function getSortKeyByLetterIndex( $index ) {
388 return $this->getFirstLetterData()['keys'][$index];
394 public function getFirstLetterCount() {
395 return count( $this->getFirstLetterData()['chars'] );
401 public static function isCjk( $codepoint ) {
402 foreach ( self
::$cjkBlocks as $block ) {
403 if ( $codepoint >= $block[0] && $codepoint <= $block[1] ) {
411 * Return the version of ICU library used by PHP's intl extension,
412 * or false when the extension is not installed of the version
413 * can't be determined.
415 * The constant INTL_ICU_VERSION this function refers to isn't really
416 * documented. It is available since PHP 5.3.7 (see PHP bug 54561).
417 * This function will return false on older PHPs.
420 * @return string|bool
422 static function getICUVersion() {
423 return defined( 'INTL_ICU_VERSION' ) ? INTL_ICU_VERSION
: false;
427 * Return the version of Unicode appropriate for the version of ICU library
428 * currently in use, or false when it can't be determined.
431 * @return string|bool
433 static function getUnicodeVersionForICU() {
434 $icuVersion = IcuCollation
::getICUVersion();
435 if ( !$icuVersion ) {
439 $versionPrefix = substr( $icuVersion, 0, 3 );
440 // Source: http://site.icu-project.org/download
454 if ( isset( $map[$versionPrefix] ) ) {
455 return $map[$versionPrefix];