3 * Database row sorting.
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.
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.
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
23 abstract class Collation
{
29 static function singleton() {
30 if ( !self
::$instance ) {
31 global $wgCategoryCollation;
32 self
::$instance = self
::factory( $wgCategoryCollation );
34 return self
::$instance;
39 * @param $collationName string
42 static function factory( $collationName ) {
43 switch( $collationName ) {
45 return new UppercaseCollation
;
47 return new IdentityCollation
;
49 return new IcuCollation( 'root' );
52 if ( preg_match( '/^uca-([a-z-]+)$/', $collationName, $match ) ) {
53 return new IcuCollation( $match[1] );
56 # Provide a mechanism for extensions to hook in.
57 $collationObject = null;
58 wfRunHooks( 'Collation::factory', array( $collationName, &$collationObject ) );
60 if ( $collationObject instanceof Collation
) {
61 return $collationObject;
64 // If all else fails...
65 throw new MWException( __METHOD__
.": unknown collation type \"$collationName\"" );
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
77 * @param string $string UTF-8 string
78 * @return string Binary sortkey
80 abstract function getSortKey( $string );
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
100 * etc., assuming for the sake of argument that $wgCapitalLinks is false.
102 * @param string $string UTF-8 string
103 * @return string UTF-8 string corresponding to the first letter of input
105 abstract function getFirstLetter( $string );
108 class UppercaseCollation
extends Collation
{
110 function __construct() {
111 // Get a language object so that we can use the generic UTF-8 uppercase
113 $this->lang
= Language
::factory( 'en' );
116 function getSortKey( $string ) {
117 return $this->lang
->uc( $string );
120 function getFirstLetter( $string ) {
121 if ( $string[0] == "\0" ) {
122 $string = substr( $string, 1 );
124 return $this->lang
->ucfirst( $this->lang
->firstChar( $string ) );
129 * Collation class that's essentially a no-op.
131 * Does sorting based on binary value of the string.
132 * Like how things were pre 1.17.
134 class IdentityCollation
extends Collation
{
136 function getSortKey( $string ) {
140 function getFirstLetter( $string ) {
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 );
147 return $wgContLang->firstChar( $string );
152 class IcuCollation
extends Collation
{
153 var $primaryCollator, $mainCollator, $locale;
154 var $firstLetterData;
157 * Unified CJK blocks.
159 * The same definition of a CJK block must be used for both Collation and
160 * generateCollationData.php. These blocks are omitted from the first
161 * letter data, as an optimisation measure and because the default UCA table
162 * is pretty useless for sorting Chinese text anyway. Japanese and Korean
163 * blocks are not included here, because they are smaller and more useful.
165 static $cjkBlocks = array(
166 array( 0x2E80, 0x2EFF ), // CJK Radicals Supplement
167 array( 0x2F00, 0x2FDF ), // Kangxi Radicals
168 array( 0x2FF0, 0x2FFF ), // Ideographic Description Characters
169 array( 0x3000, 0x303F ), // CJK Symbols and Punctuation
170 array( 0x31C0, 0x31EF ), // CJK Strokes
171 array( 0x3200, 0x32FF ), // Enclosed CJK Letters and Months
172 array( 0x3300, 0x33FF ), // CJK Compatibility
173 array( 0x3400, 0x4DBF ), // CJK Unified Ideographs Extension A
174 array( 0x4E00, 0x9FFF ), // CJK Unified Ideographs
175 array( 0xF900, 0xFAFF ), // CJK Compatibility Ideographs
176 array( 0xFE30, 0xFE4F ), // CJK Compatibility Forms
177 array( 0x20000, 0x2A6DF ), // CJK Unified Ideographs Extension B
178 array( 0x2A700, 0x2B73F ), // CJK Unified Ideographs Extension C
179 array( 0x2B740, 0x2B81F ), // CJK Unified Ideographs Extension D
180 array( 0x2F800, 0x2FA1F ), // CJK Compatibility Ideographs Supplement
184 * Additional characters (or character groups) to be considered first-letters
186 * Generated based on the primary level of Unicode collation tailorings
187 * available at http://developer.mimer.com/charts/tailorings.htm .
189 * Empty arrays are intended; this signifies that the data for the language is
190 * available and that there are, in fact, no additional letters to consider.
192 static $tailoringFirstLetters = array(
193 // Verified by native speakers
194 'pl' => array( "Ą", "Ć", "Ę", "Ł", "Ń", "Ó", "Ś", "Ź", "Ż" ),
195 'fi' => array( "Å", "Ä", "Ö" ),
196 // Not verified, but likely correct
198 'ast' => array( "CH", "LL", "Ñ" ),
199 'az' => array( "Ç", "Ə", "Ğ", "İ", "Ö", "Ş", "Ü" ),
200 'be' => array( "Ё" ),
202 'br' => array( "CH", "C'H" ),
203 'bs' => array( "Č", "Ć", "DŽ", "Đ", "LJ", "NJ", "Š", "Ž" ),
206 'cs' => array( "Č", "CH", "Ř", "Š", "Ž" ),
207 'cy' => array( "CH", "DD", "FF", "NG", "LL", "PH", "RH", "TH" ),
208 'da' => array( "Æ", "Ø", "Å" ),
210 'dsb' => array( "Č", "Ć", "DŹ", "Ě", "CH", "Ł", "Ń", "Ŕ", "Š", "Ś", "Ž", "Ź" ),
213 'eo' => array( "Ĉ", "Ĝ", "Ĥ", "Ĵ", "Ŝ", "Ŭ" ),
214 'es' => array( "Ñ" ),
215 'et' => array( "Š", "Ž", "Õ", "Ä", "Ö", "Ü" ),
216 'eu' => array( "Ñ" ),
217 'fo' => array( "Á", "Ð", "Í", "Ó", "Ú", "Ý", "Æ", "Ø", "Å" ),
219 'fur' => array( "À", "Á", "Â", "È", "Ì", "Ò", "Ù" ),
223 'gl' => array( "CH", "LL", "Ñ" ),
224 'hr' => array( "Č", "Ć", "DŽ", "Đ", "LJ", "NJ", "Š", "Ž" ),
225 'hsb' => array( "Č", "DŹ", "Ě", "CH", "Ł", "Ń", "Ř", "Š", "Ć", "Ž" ),
226 'hu' => array( "CS", "DZ", "DZS", "GY", "LY", "NY", "Ö", "SZ", "TY", "Ü", "ZS" ),
227 'is' => array( "Á", "Ð", "É", "Í", "Ó", "Ú", "Ý", "Þ", "Æ", "Ö", "Å" ),
229 'kk' => array( "Ү", "І" ),
230 'kl' => array( "Æ", "Ø", "Å" ),
231 'ku' => array( "Ç", "Ê", "Î", "Ş", "Û" ),
232 'ky' => array( "Ё" ),
235 'lt' => array( "Č", "Š", "Ž" ),
236 'lv' => array( "Č", "Ģ", "Ķ", "Ļ", "Ņ", "Š", "Ž" ),
238 'mo' => array( "Ă", "Â", "Î", "Ş", "Ţ" ),
239 'mt' => array( "Ċ", "Ġ", "GĦ", "Ħ", "Ż" ),
241 'no' => array( "Æ", "Ø", "Å" ),
245 'ro' => array( "Ă", "Â", "Î", "Ş", "Ţ" ),
247 'rup' => array( "Ă", "Â", "Î", "Ľ", "Ń", "Ş", "Ţ" ),
249 'sk' => array( "Ä", "Č", "CH", "Ô", "Š", "Ž" ),
250 'sl' => array( "Č", "Š", "Ž" ),
251 'smn' => array( "Á", "Č", "Đ", "Ŋ", "Š", "Ŧ", "Ž", "Æ", "Ø", "Å", "Ä", "Ö" ),
252 'sq' => array( "Ç", "DH", "Ë", "GJ", "LL", "NJ", "RR", "SH", "TH", "XH", "ZH" ),
254 'sv' => array( "Å", "Ä", "Ö" ),
255 'tk' => array( "Ç", "Ä", "Ž", "Ň", "Ö", "Ş", "Ü", "Ý" ),
256 'tl' => array( "Ñ", "NG" ), /* 'fil' in the data source */
257 'tr' => array( "Ç", "Ğ", "İ", "Ö", "Ş", "Ü" ),
258 'tt' => array( "Ә", "Ө", "Ү", "Җ", "Ң", "Һ" ),
259 'uk' => array( "Ґ", "Ь" ),
260 'uz' => array( "CH", "G'", "NG", "O'", "SH" ),
261 'vi' => array( "Ă", "Â", "Đ", "Ê", "Ô", "Ơ", "Ư" ),
264 const RECORD_LENGTH
= 14;
266 function __construct( $locale ) {
267 if ( !extension_loaded( 'intl' ) ) {
268 throw new MWException( 'An ICU collation was requested, ' .
269 'but the intl extension is not available.' );
271 $this->locale
= $locale;
272 $this->mainCollator
= Collator
::create( $locale );
273 if ( !$this->mainCollator
) {
274 throw new MWException( "Invalid ICU locale specified for collation: $locale" );
277 $this->primaryCollator
= Collator
::create( $locale );
278 $this->primaryCollator
->setStrength( Collator
::PRIMARY
);
281 function getSortKey( $string ) {
282 // intl extension produces non null-terminated
283 // strings. Appending '' fixes it so that it doesn't generate
284 // a warning on each access in debug php.
285 wfSuppressWarnings();
286 $key = $this->mainCollator
->getSortKey( $string ) . '';
291 function getPrimarySortKey( $string ) {
292 wfSuppressWarnings();
293 $key = $this->primaryCollator
->getSortKey( $string ) . '';
298 function getFirstLetter( $string ) {
299 $string = strval( $string );
300 if ( $string === '' ) {
305 $firstChar = mb_substr( $string, 0, 1, 'UTF-8' );
306 if ( ord( $firstChar ) > 0x7f
307 && self
::isCjk( utf8ToCodepoint( $firstChar ) ) )
312 $sortKey = $this->getPrimarySortKey( $string );
314 // Do a binary search to find the correct letter to sort under
315 $min = $this->findLowerBound(
316 array( $this, 'getSortKeyByLetterIndex' ),
317 $this->getFirstLetterCount(),
321 if ( $min === false ) {
322 // Before the first letter
325 return $this->getLetterByIndex( $min );
328 function getFirstLetterData() {
329 if ( $this->firstLetterData
!== null ) {
330 return $this->firstLetterData
;
333 $cache = wfGetCache( CACHE_ANYTHING
);
334 $cacheKey = wfMemcKey( 'first-letters', $this->locale
);
335 $cacheEntry = $cache->get( $cacheKey );
338 $this->firstLetterData
= $cacheEntry;
339 return $this->firstLetterData
;
342 // Generate data from serialized data file
344 if ( isset ( self
::$tailoringFirstLetters[$this->locale
] ) ) {
345 $letters = wfGetPrecompiledData( "first-letters-root.ser" );
346 $letters = array_merge( $letters, self
::$tailoringFirstLetters[$this->locale
] );
348 $letters = wfGetPrecompiledData( "first-letters-{$this->locale}.ser" );
349 if ( $letters === false ) {
350 throw new MWException( "MediaWiki does not support ICU locale " .
351 "\"{$this->locale}\"" );
357 // It's impossible to have the precompiled data file properly sorted,
358 // because the sort order changes depending on ICU version. If the
359 // array is not properly sorted, the binary search will return random
362 // We also take this opportunity to remove primary collisions.
363 $letterMap = array();
364 foreach ( $letters as $letter ) {
365 $key = $this->getPrimarySortKey( $letter );
366 if ( isset( $letterMap[$key] ) ) {
368 // Keep whichever one sorts first in the main collator
369 if ( $this->mainCollator
->compare( $letter, $letterMap[$key] ) < 0 ) {
370 $letterMap[$key] = $letter;
373 $letterMap[$key] = $letter;
376 ksort( $letterMap, SORT_STRING
);
378 'chars' => array_values( $letterMap ),
379 'keys' => array_keys( $letterMap )
382 // Reduce memory usage before caching
386 $this->firstLetterData
= $data;
387 $cache->set( $cacheKey, $data, 86400 * 7 /* 1 week */ );
391 function getLetterByIndex( $index ) {
392 if ( $this->firstLetterData
=== null ) {
393 $this->getFirstLetterData();
395 return $this->firstLetterData
['chars'][$index];
398 function getSortKeyByLetterIndex( $index ) {
399 if ( $this->firstLetterData
=== null ) {
400 $this->getFirstLetterData();
402 return $this->firstLetterData
['keys'][$index];
405 function getFirstLetterCount() {
406 if ( $this->firstLetterData
=== null ) {
407 $this->getFirstLetterData();
409 return count( $this->firstLetterData
['chars'] );
413 * Do a binary search, and return the index of the largest item that sorts
414 * less than or equal to the target value.
416 * @param $valueCallback array A function to call to get the value with
417 * a given array index.
418 * @param $valueCount int The number of items accessible via $valueCallback,
419 * indexed from 0 to $valueCount - 1
420 * @param $comparisonCallback array A callback to compare two values, returning
421 * -1, 0 or 1 in the style of strcmp().
422 * @param $target string The target value to find.
424 * @return int|bool The item index of the lower bound, or false if the target value
425 * sorts before all items.
427 function findLowerBound( $valueCallback, $valueCount, $comparisonCallback, $target ) {
428 if ( $valueCount === 0 ) {
435 $mid = $min +
( ( $max - $min ) >> 1 );
436 $item = call_user_func( $valueCallback, $mid );
437 $comparison = call_user_func( $comparisonCallback, $target, $item );
438 if ( $comparison > 0 ) {
440 } elseif ( $comparison == 0 ) {
446 } while ( $min < $max - 1 );
449 $item = call_user_func( $valueCallback, $min );
450 $comparison = call_user_func( $comparisonCallback, $target, $item );
451 if ( $comparison < 0 ) {
452 // Before the first item
459 static function isCjk( $codepoint ) {
460 foreach ( self
::$cjkBlocks as $block ) {
461 if ( $codepoint >= $block[0] && $codepoint <= $block[1] ) {
469 * Return the version of ICU library used by PHP's intl extension,
470 * or false when the extension is not installed of the version
471 * can't be determined.
473 * The constant INTL_ICU_VERSION this function refers to isn't really
474 * documented. It is available since PHP 5.3.7 (see PHP bug 54561).
475 * This function will return false on older PHPs.
478 * @return string|false
480 static function getICUVersion() {
481 return defined( 'INTL_ICU_VERSION' ) ? INTL_ICU_VERSION
: false;
485 * Return the version of Unicode appropriate for the version of ICU library
486 * currently in use, or false when it can't be determined.
489 * @return string|false
491 static function getUnicodeVersionForICU() {
492 $icuVersion = IcuCollation
::getICUVersion();
493 if ( !$icuVersion ) {
497 $versionPrefix = substr( $icuVersion, 0, 3 );
498 // Source: http://site.icu-project.org/download
512 if ( isset( $map[$versionPrefix] ) ) {
513 return $map[$versionPrefix];