Plural rules: updates for UTS #35 Rev 33
[lhc/web/wiklou.git] / includes / cache / LocalisationCache.php
1 <?php
2 /**
3 * Cache of the contents of localisation files.
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 define( 'MW_LC_VERSION', 2 );
24
25 /**
26 * Class for caching the contents of localisation files, Messages*.php
27 * and *.i18n.php.
28 *
29 * An instance of this class is available using Language::getLocalisationCache().
30 *
31 * The values retrieved from here are merged, containing items from extension
32 * files, core messages files and the language fallback sequence (e.g. zh-cn ->
33 * zh-hans -> en ). Some common errors are corrected, for example namespace
34 * names with spaces instead of underscores, but heavyweight processing, such
35 * as grammatical transformation, is done by the caller.
36 */
37 class LocalisationCache {
38 /** Configuration associative array */
39 private $conf;
40
41 /**
42 * True if recaching should only be done on an explicit call to recache().
43 * Setting this reduces the overhead of cache freshness checking, which
44 * requires doing a stat() for every extension i18n file.
45 */
46 private $manualRecache = false;
47
48 /**
49 * True to treat all files as expired until they are regenerated by this object.
50 */
51 private $forceRecache = false;
52
53 /**
54 * The cache data. 3-d array, where the first key is the language code,
55 * the second key is the item key e.g. 'messages', and the third key is
56 * an item specific subkey index. Some items are not arrays and so for those
57 * items, there are no subkeys.
58 */
59 protected $data = array();
60
61 /**
62 * The persistent store object. An instance of LCStore.
63 *
64 * @var LCStore
65 */
66 private $store;
67
68 /**
69 * A 2-d associative array, code/key, where presence indicates that the item
70 * is loaded. Value arbitrary.
71 *
72 * For split items, if set, this indicates that all of the subitems have been
73 * loaded.
74 */
75 private $loadedItems = array();
76
77 /**
78 * A 3-d associative array, code/key/subkey, where presence indicates that
79 * the subitem is loaded. Only used for the split items, i.e. messages.
80 */
81 private $loadedSubitems = array();
82
83 /**
84 * An array where presence of a key indicates that that language has been
85 * initialised. Initialisation includes checking for cache expiry and doing
86 * any necessary updates.
87 */
88 private $initialisedLangs = array();
89
90 /**
91 * An array mapping non-existent pseudo-languages to fallback languages. This
92 * is filled by initShallowFallback() when data is requested from a language
93 * that lacks a Messages*.php file.
94 */
95 private $shallowFallbacks = array();
96
97 /**
98 * An array where the keys are codes that have been recached by this instance.
99 */
100 private $recachedLangs = array();
101
102 /**
103 * All item keys
104 */
105 static public $allKeys = array(
106 'fallback', 'namespaceNames', 'bookstoreList',
107 'magicWords', 'messages', 'rtl', 'capitalizeAllNouns', 'digitTransformTable',
108 'separatorTransformTable', 'fallback8bitEncoding', 'linkPrefixExtension',
109 'linkTrail', 'linkPrefixCharset', 'namespaceAliases',
110 'dateFormats', 'datePreferences', 'datePreferenceMigrationMap',
111 'defaultDateFormat', 'extraUserToggles', 'specialPageAliases',
112 'imageFiles', 'preloadedMessages', 'namespaceGenderAliases',
113 'digitGroupingPattern', 'pluralRules', 'pluralRuleTypes', 'compiledPluralRules',
114 );
115
116 /**
117 * Keys for items which consist of associative arrays, which may be merged
118 * by a fallback sequence.
119 */
120 static public $mergeableMapKeys = array( 'messages', 'namespaceNames',
121 'dateFormats', 'imageFiles', 'preloadedMessages'
122 );
123
124 /**
125 * Keys for items which are a numbered array.
126 */
127 static public $mergeableListKeys = array( 'extraUserToggles' );
128
129 /**
130 * Keys for items which contain an array of arrays of equivalent aliases
131 * for each subitem. The aliases may be merged by a fallback sequence.
132 */
133 static public $mergeableAliasListKeys = array( 'specialPageAliases' );
134
135 /**
136 * Keys for items which contain an associative array, and may be merged if
137 * the primary value contains the special array key "inherit". That array
138 * key is removed after the first merge.
139 */
140 static public $optionalMergeKeys = array( 'bookstoreList' );
141
142 /**
143 * Keys for items that are formatted like $magicWords
144 */
145 static public $magicWordKeys = array( 'magicWords' );
146
147 /**
148 * Keys for items where the subitems are stored in the backend separately.
149 */
150 static public $splitKeys = array( 'messages' );
151
152 /**
153 * Keys which are loaded automatically by initLanguage()
154 */
155 static public $preloadedKeys = array( 'dateFormats', 'namespaceNames' );
156
157 /**
158 * Associative array of cached plural rules. The key is the language code,
159 * the value is an array of plural rules for that language.
160 */
161 private $pluralRules = null;
162
163 /**
164 * Associative array of cached plural rule types. The key is the language
165 * code, the value is an array of plural rule types for that language. For
166 * example, $pluralRuleTypes['ar'] = ['zero', 'one', 'two', 'few', 'many'].
167 * The index for each rule type matches the index for the rule in
168 * $pluralRules, thus allowing correlation between the two. The reason we
169 * don't just use the type names as the keys in $pluralRules is because
170 * Language::convertPlural applies the rules based on numeric order (or
171 * explicit numeric parameter), not based on the name of the rule type. For
172 * example, {{plural:count|wordform1|wordform2|wordform3}}, rather than
173 * {{plural:count|one=wordform1|two=wordform2|many=wordform3}}.
174 */
175 private $pluralRuleTypes = null;
176
177 private $mergeableKeys = null;
178
179 /**
180 * Constructor.
181 * For constructor parameters, see the documentation in DefaultSettings.php
182 * for $wgLocalisationCacheConf.
183 *
184 * @param $conf Array
185 * @throws MWException
186 */
187 function __construct( $conf ) {
188 global $wgCacheDirectory;
189
190 $this->conf = $conf;
191 $storeConf = array();
192 if ( !empty( $conf['storeClass'] ) ) {
193 $storeClass = $conf['storeClass'];
194 } else {
195 switch ( $conf['store'] ) {
196 case 'files':
197 case 'file':
198 $storeClass = 'LCStoreCDB';
199 break;
200 case 'db':
201 $storeClass = 'LCStoreDB';
202 break;
203 case 'accel':
204 $storeClass = 'LCStoreAccel';
205 break;
206 case 'detect':
207 $storeClass = $wgCacheDirectory ? 'LCStoreCDB' : 'LCStoreDB';
208 break;
209 default:
210 throw new MWException(
211 'Please set $wgLocalisationCacheConf[\'store\'] to something sensible.' );
212 }
213 }
214
215 wfDebug( get_class( $this ) . ": using store $storeClass\n" );
216 if ( !empty( $conf['storeDirectory'] ) ) {
217 $storeConf['directory'] = $conf['storeDirectory'];
218 }
219
220 $this->store = new $storeClass( $storeConf );
221 foreach ( array( 'manualRecache', 'forceRecache' ) as $var ) {
222 if ( isset( $conf[$var] ) ) {
223 $this->$var = $conf[$var];
224 }
225 }
226 }
227
228 /**
229 * Returns true if the given key is mergeable, that is, if it is an associative
230 * array which can be merged through a fallback sequence.
231 * @param $key
232 * @return bool
233 */
234 public function isMergeableKey( $key ) {
235 if ( $this->mergeableKeys === null ) {
236 $this->mergeableKeys = array_flip( array_merge(
237 self::$mergeableMapKeys,
238 self::$mergeableListKeys,
239 self::$mergeableAliasListKeys,
240 self::$optionalMergeKeys,
241 self::$magicWordKeys
242 ) );
243 }
244
245 return isset( $this->mergeableKeys[$key] );
246 }
247
248 /**
249 * Get a cache item.
250 *
251 * Warning: this may be slow for split items (messages), since it will
252 * need to fetch all of the subitems from the cache individually.
253 * @param $code
254 * @param $key
255 * @return mixed
256 */
257 public function getItem( $code, $key ) {
258 if ( !isset( $this->loadedItems[$code][$key] ) ) {
259 wfProfileIn( __METHOD__ . '-load' );
260 $this->loadItem( $code, $key );
261 wfProfileOut( __METHOD__ . '-load' );
262 }
263
264 if ( $key === 'fallback' && isset( $this->shallowFallbacks[$code] ) ) {
265 return $this->shallowFallbacks[$code];
266 }
267
268 return $this->data[$code][$key];
269 }
270
271 /**
272 * Get a subitem, for instance a single message for a given language.
273 * @param $code
274 * @param $key
275 * @param $subkey
276 * @return null
277 */
278 public function getSubitem( $code, $key, $subkey ) {
279 if ( !isset( $this->loadedSubitems[$code][$key][$subkey] ) &&
280 !isset( $this->loadedItems[$code][$key] )
281 ) {
282 wfProfileIn( __METHOD__ . '-load' );
283 $this->loadSubitem( $code, $key, $subkey );
284 wfProfileOut( __METHOD__ . '-load' );
285 }
286
287 if ( isset( $this->data[$code][$key][$subkey] ) ) {
288 return $this->data[$code][$key][$subkey];
289 } else {
290 return null;
291 }
292 }
293
294 /**
295 * Get the list of subitem keys for a given item.
296 *
297 * This is faster than array_keys($lc->getItem(...)) for the items listed in
298 * self::$splitKeys.
299 *
300 * Will return null if the item is not found, or false if the item is not an
301 * array.
302 * @param $code
303 * @param $key
304 * @return bool|null|string
305 */
306 public function getSubitemList( $code, $key ) {
307 if ( in_array( $key, self::$splitKeys ) ) {
308 return $this->getSubitem( $code, 'list', $key );
309 } else {
310 $item = $this->getItem( $code, $key );
311 if ( is_array( $item ) ) {
312 return array_keys( $item );
313 } else {
314 return false;
315 }
316 }
317 }
318
319 /**
320 * Load an item into the cache.
321 * @param $code
322 * @param $key
323 */
324 protected function loadItem( $code, $key ) {
325 if ( !isset( $this->initialisedLangs[$code] ) ) {
326 $this->initLanguage( $code );
327 }
328
329 // Check to see if initLanguage() loaded it for us
330 if ( isset( $this->loadedItems[$code][$key] ) ) {
331 return;
332 }
333
334 if ( isset( $this->shallowFallbacks[$code] ) ) {
335 $this->loadItem( $this->shallowFallbacks[$code], $key );
336
337 return;
338 }
339
340 if ( in_array( $key, self::$splitKeys ) ) {
341 $subkeyList = $this->getSubitem( $code, 'list', $key );
342 foreach ( $subkeyList as $subkey ) {
343 if ( isset( $this->data[$code][$key][$subkey] ) ) {
344 continue;
345 }
346 $this->data[$code][$key][$subkey] = $this->getSubitem( $code, $key, $subkey );
347 }
348 } else {
349 $this->data[$code][$key] = $this->store->get( $code, $key );
350 }
351
352 $this->loadedItems[$code][$key] = true;
353 }
354
355 /**
356 * Load a subitem into the cache
357 * @param $code
358 * @param $key
359 * @param $subkey
360 */
361 protected function loadSubitem( $code, $key, $subkey ) {
362 if ( !in_array( $key, self::$splitKeys ) ) {
363 $this->loadItem( $code, $key );
364
365 return;
366 }
367
368 if ( !isset( $this->initialisedLangs[$code] ) ) {
369 $this->initLanguage( $code );
370 }
371
372 // Check to see if initLanguage() loaded it for us
373 if ( isset( $this->loadedItems[$code][$key] ) ||
374 isset( $this->loadedSubitems[$code][$key][$subkey] )
375 ) {
376 return;
377 }
378
379 if ( isset( $this->shallowFallbacks[$code] ) ) {
380 $this->loadSubitem( $this->shallowFallbacks[$code], $key, $subkey );
381
382 return;
383 }
384
385 $value = $this->store->get( $code, "$key:$subkey" );
386 $this->data[$code][$key][$subkey] = $value;
387 $this->loadedSubitems[$code][$key][$subkey] = true;
388 }
389
390 /**
391 * Returns true if the cache identified by $code is missing or expired.
392 *
393 * @param string $code
394 *
395 * @return bool
396 */
397 public function isExpired( $code ) {
398 if ( $this->forceRecache && !isset( $this->recachedLangs[$code] ) ) {
399 wfDebug( __METHOD__ . "($code): forced reload\n" );
400
401 return true;
402 }
403
404 $deps = $this->store->get( $code, 'deps' );
405 $keys = $this->store->get( $code, 'list' );
406 $preload = $this->store->get( $code, 'preload' );
407 // Different keys may expire separately, at least in LCStoreAccel
408 if ( $deps === null || $keys === null || $preload === null ) {
409 wfDebug( __METHOD__ . "($code): cache missing, need to make one\n" );
410
411 return true;
412 }
413
414 foreach ( $deps as $dep ) {
415 // Because we're unserializing stuff from cache, we
416 // could receive objects of classes that don't exist
417 // anymore (e.g. uninstalled extensions)
418 // When this happens, always expire the cache
419 if ( !$dep instanceof CacheDependency || $dep->isExpired() ) {
420 wfDebug( __METHOD__ . "($code): cache for $code expired due to " .
421 get_class( $dep ) . "\n" );
422
423 return true;
424 }
425 }
426
427 return false;
428 }
429
430 /**
431 * Initialise a language in this object. Rebuild the cache if necessary.
432 * @param $code
433 * @throws MWException
434 */
435 protected function initLanguage( $code ) {
436 if ( isset( $this->initialisedLangs[$code] ) ) {
437 return;
438 }
439
440 $this->initialisedLangs[$code] = true;
441
442 # If the code is of the wrong form for a Messages*.php file, do a shallow fallback
443 if ( !Language::isValidBuiltInCode( $code ) ) {
444 $this->initShallowFallback( $code, 'en' );
445
446 return;
447 }
448
449 # Recache the data if necessary
450 if ( !$this->manualRecache && $this->isExpired( $code ) ) {
451 if ( file_exists( Language::getMessagesFileName( $code ) ) ) {
452 $this->recache( $code );
453 } elseif ( $code === 'en' ) {
454 throw new MWException( 'MessagesEn.php is missing.' );
455 } else {
456 $this->initShallowFallback( $code, 'en' );
457 }
458
459 return;
460 }
461
462 # Preload some stuff
463 $preload = $this->getItem( $code, 'preload' );
464 if ( $preload === null ) {
465 if ( $this->manualRecache ) {
466 // No Messages*.php file. Do shallow fallback to en.
467 if ( $code === 'en' ) {
468 throw new MWException( 'No localisation cache found for English. ' .
469 'Please run maintenance/rebuildLocalisationCache.php.' );
470 }
471 $this->initShallowFallback( $code, 'en' );
472
473 return;
474 } else {
475 throw new MWException( 'Invalid or missing localisation cache.' );
476 }
477 }
478 $this->data[$code] = $preload;
479 foreach ( $preload as $key => $item ) {
480 if ( in_array( $key, self::$splitKeys ) ) {
481 foreach ( $item as $subkey => $subitem ) {
482 $this->loadedSubitems[$code][$key][$subkey] = true;
483 }
484 } else {
485 $this->loadedItems[$code][$key] = true;
486 }
487 }
488 }
489
490 /**
491 * Create a fallback from one language to another, without creating a
492 * complete persistent cache.
493 * @param $primaryCode
494 * @param $fallbackCode
495 */
496 public function initShallowFallback( $primaryCode, $fallbackCode ) {
497 $this->data[$primaryCode] =& $this->data[$fallbackCode];
498 $this->loadedItems[$primaryCode] =& $this->loadedItems[$fallbackCode];
499 $this->loadedSubitems[$primaryCode] =& $this->loadedSubitems[$fallbackCode];
500 $this->shallowFallbacks[$primaryCode] = $fallbackCode;
501 }
502
503 /**
504 * Read a PHP file containing localisation data.
505 * @param $_fileName
506 * @param $_fileType
507 * @throws MWException
508 * @return array
509 */
510 protected function readPHPFile( $_fileName, $_fileType ) {
511 wfProfileIn( __METHOD__ );
512 // Disable APC caching
513 $_apcEnabled = ini_set( 'apc.cache_by_default', '0' );
514 include $_fileName;
515 ini_set( 'apc.cache_by_default', $_apcEnabled );
516
517 if ( $_fileType == 'core' || $_fileType == 'extension' ) {
518 $data = compact( self::$allKeys );
519 } elseif ( $_fileType == 'aliases' ) {
520 $data = compact( 'aliases' );
521 } else {
522 wfProfileOut( __METHOD__ );
523 throw new MWException( __METHOD__ . ": Invalid file type: $_fileType" );
524 }
525 wfProfileOut( __METHOD__ );
526
527 return $data;
528 }
529
530 /**
531 * Get the compiled plural rules for a given language from the XML files.
532 * @since 1.20
533 */
534 public function getCompiledPluralRules( $code ) {
535 $rules = $this->getPluralRules( $code );
536 if ( $rules === null ) {
537 return null;
538 }
539 try {
540 $compiledRules = CLDRPluralRuleEvaluator::compile( $rules );
541 } catch ( CLDRPluralRuleError $e ) {
542 wfDebugLog( 'l10n', $e->getMessage() . "\n" );
543
544 return array();
545 }
546
547 return $compiledRules;
548 }
549
550 /**
551 * Get the plural rules for a given language from the XML files.
552 * Cached.
553 * @since 1.20
554 */
555 public function getPluralRules( $code ) {
556 if ( $this->pluralRules === null ) {
557 $this->loadPluralFiles();
558 }
559 if ( !isset( $this->pluralRules[$code] ) ) {
560 return null;
561 } else {
562 return $this->pluralRules[$code];
563 }
564 }
565
566 /**
567 * Get the plural rule types for a given language from the XML files.
568 * Cached.
569 * @since 1.22
570 */
571 public function getPluralRuleTypes( $code ) {
572 if ( $this->pluralRuleTypes === null ) {
573 $this->loadPluralFiles();
574 }
575 if ( !isset( $this->pluralRuleTypes[$code] ) ) {
576 return null;
577 } else {
578 return $this->pluralRuleTypes[$code];
579 }
580 }
581
582 /**
583 * Load the plural XML files.
584 */
585 protected function loadPluralFiles() {
586 global $IP;
587 $cldrPlural = "$IP/languages/data/plurals.xml";
588 $mwPlural = "$IP/languages/data/plurals-mediawiki.xml";
589 // Load CLDR plural rules
590 $this->loadPluralFile( $cldrPlural );
591 if ( file_exists( $mwPlural ) ) {
592 // Override or extend
593 $this->loadPluralFile( $mwPlural );
594 }
595 }
596
597 /**
598 * Load a plural XML file with the given filename, compile the relevant
599 * rules, and save the compiled rules in a process-local cache.
600 */
601 protected function loadPluralFile( $fileName ) {
602 $doc = new DOMDocument;
603 $doc->load( $fileName );
604 $rulesets = $doc->getElementsByTagName( "pluralRules" );
605 foreach ( $rulesets as $ruleset ) {
606 $codes = $ruleset->getAttribute( 'locales' );
607 $rules = array();
608 $ruleTypes = array();
609 $ruleElements = $ruleset->getElementsByTagName( "pluralRule" );
610 foreach ( $ruleElements as $elt ) {
611 $ruleType = $elt->getAttribute( 'count' );
612 if ( $ruleType === 'other' ) {
613 // Don't record "other" rules, which have an empty condition
614 continue;
615 }
616 $rules[] = $elt->nodeValue;
617 $ruleTypes[] = $ruleType;
618 }
619 foreach ( explode( ' ', $codes ) as $code ) {
620 $this->pluralRules[$code] = $rules;
621 $this->pluralRuleTypes[$code] = $ruleTypes;
622 }
623 }
624 }
625
626 /**
627 * Read the data from the source files for a given language, and register
628 * the relevant dependencies in the $deps array. If the localisation
629 * exists, the data array is returned, otherwise false is returned.
630 */
631 protected function readSourceFilesAndRegisterDeps( $code, &$deps ) {
632 global $IP;
633 wfProfileIn( __METHOD__ );
634
635 $fileName = Language::getMessagesFileName( $code );
636 if ( !file_exists( $fileName ) ) {
637 wfProfileOut( __METHOD__ );
638
639 return false;
640 }
641
642 $deps[] = new FileDependency( $fileName );
643 $data = $this->readPHPFile( $fileName, 'core' );
644
645 # Load CLDR plural rules for JavaScript
646 $data['pluralRules'] = $this->getPluralRules( $code );
647 # And for PHP
648 $data['compiledPluralRules'] = $this->getCompiledPluralRules( $code );
649 # Load plural rule types
650 $data['pluralRuleTypes'] = $this->getPluralRuleTypes( $code );
651
652 $deps['plurals'] = new FileDependency( "$IP/languages/data/plurals.xml" );
653 $deps['plurals-mw'] = new FileDependency( "$IP/languages/data/plurals-mediawiki.xml" );
654
655 wfProfileOut( __METHOD__ );
656
657 return $data;
658 }
659
660 /**
661 * Merge two localisation values, a primary and a fallback, overwriting the
662 * primary value in place.
663 * @param $key
664 * @param $value
665 * @param $fallbackValue
666 */
667 protected function mergeItem( $key, &$value, $fallbackValue ) {
668 if ( !is_null( $value ) ) {
669 if ( !is_null( $fallbackValue ) ) {
670 if ( in_array( $key, self::$mergeableMapKeys ) ) {
671 $value = $value + $fallbackValue;
672 } elseif ( in_array( $key, self::$mergeableListKeys ) ) {
673 $value = array_unique( array_merge( $fallbackValue, $value ) );
674 } elseif ( in_array( $key, self::$mergeableAliasListKeys ) ) {
675 $value = array_merge_recursive( $value, $fallbackValue );
676 } elseif ( in_array( $key, self::$optionalMergeKeys ) ) {
677 if ( !empty( $value['inherit'] ) ) {
678 $value = array_merge( $fallbackValue, $value );
679 }
680
681 if ( isset( $value['inherit'] ) ) {
682 unset( $value['inherit'] );
683 }
684 } elseif ( in_array( $key, self::$magicWordKeys ) ) {
685 $this->mergeMagicWords( $value, $fallbackValue );
686 }
687 }
688 } else {
689 $value = $fallbackValue;
690 }
691 }
692
693 /**
694 * @param $value
695 * @param $fallbackValue
696 */
697 protected function mergeMagicWords( &$value, $fallbackValue ) {
698 foreach ( $fallbackValue as $magicName => $fallbackInfo ) {
699 if ( !isset( $value[$magicName] ) ) {
700 $value[$magicName] = $fallbackInfo;
701 } else {
702 $oldSynonyms = array_slice( $fallbackInfo, 1 );
703 $newSynonyms = array_slice( $value[$magicName], 1 );
704 $synonyms = array_values( array_unique( array_merge(
705 $newSynonyms, $oldSynonyms ) ) );
706 $value[$magicName] = array_merge( array( $fallbackInfo[0] ), $synonyms );
707 }
708 }
709 }
710
711 /**
712 * Given an array mapping language code to localisation value, such as is
713 * found in extension *.i18n.php files, iterate through a fallback sequence
714 * to merge the given data with an existing primary value.
715 *
716 * Returns true if any data from the extension array was used, false
717 * otherwise.
718 * @param $codeSequence
719 * @param $key
720 * @param $value
721 * @param $fallbackValue
722 * @return bool
723 */
724 protected function mergeExtensionItem( $codeSequence, $key, &$value, $fallbackValue ) {
725 $used = false;
726 foreach ( $codeSequence as $code ) {
727 if ( isset( $fallbackValue[$code] ) ) {
728 $this->mergeItem( $key, $value, $fallbackValue[$code] );
729 $used = true;
730 }
731 }
732
733 return $used;
734 }
735
736 /**
737 * Load localisation data for a given language for both core and extensions
738 * and save it to the persistent cache store and the process cache
739 * @param $code
740 * @throws MWException
741 */
742 public function recache( $code ) {
743 global $wgExtensionMessagesFiles;
744 wfProfileIn( __METHOD__ );
745
746 if ( !$code ) {
747 wfProfileOut( __METHOD__ );
748 throw new MWException( "Invalid language code requested" );
749 }
750 $this->recachedLangs[$code] = true;
751
752 # Initial values
753 $initialData = array_combine(
754 self::$allKeys,
755 array_fill( 0, count( self::$allKeys ), null ) );
756 $coreData = $initialData;
757 $deps = array();
758
759 # Load the primary localisation from the source file
760 $data = $this->readSourceFilesAndRegisterDeps( $code, $deps );
761 if ( $data === false ) {
762 wfDebug( __METHOD__ . ": no localisation file for $code, using fallback to en\n" );
763 $coreData['fallback'] = 'en';
764 } else {
765 wfDebug( __METHOD__ . ": got localisation for $code from source\n" );
766
767 # Merge primary localisation
768 foreach ( $data as $key => $value ) {
769 $this->mergeItem( $key, $coreData[$key], $value );
770 }
771 }
772
773 # Fill in the fallback if it's not there already
774 if ( is_null( $coreData['fallback'] ) ) {
775 $coreData['fallback'] = $code === 'en' ? false : 'en';
776 }
777 if ( $coreData['fallback'] === false ) {
778 $coreData['fallbackSequence'] = array();
779 } else {
780 $coreData['fallbackSequence'] = array_map( 'trim', explode( ',', $coreData['fallback'] ) );
781 $len = count( $coreData['fallbackSequence'] );
782
783 # Ensure that the sequence ends at en
784 if ( $coreData['fallbackSequence'][$len - 1] !== 'en' ) {
785 $coreData['fallbackSequence'][] = 'en';
786 }
787
788 # Load the fallback localisation item by item and merge it
789 foreach ( $coreData['fallbackSequence'] as $fbCode ) {
790 # Load the secondary localisation from the source file to
791 # avoid infinite cycles on cyclic fallbacks
792 $fbData = $this->readSourceFilesAndRegisterDeps( $fbCode, $deps );
793 if ( $fbData === false ) {
794 continue;
795 }
796
797 foreach ( self::$allKeys as $key ) {
798 if ( !isset( $fbData[$key] ) ) {
799 continue;
800 }
801
802 if ( is_null( $coreData[$key] ) || $this->isMergeableKey( $key ) ) {
803 $this->mergeItem( $key, $coreData[$key], $fbData[$key] );
804 }
805 }
806 }
807 }
808
809 $codeSequence = array_merge( array( $code ), $coreData['fallbackSequence'] );
810
811 # Load the extension localisations
812 # This is done after the core because we know the fallback sequence now.
813 # But it has a higher precedence for merging so that we can support things
814 # like site-specific message overrides.
815 wfProfileIn( __METHOD__ . '-extensions' );
816 $allData = $initialData;
817 foreach ( $wgExtensionMessagesFiles as $fileName ) {
818 $data = $this->readPHPFile( $fileName, 'extension' );
819 $used = false;
820
821 foreach ( $data as $key => $item ) {
822 if ( $this->mergeExtensionItem( $codeSequence, $key, $allData[$key], $item ) ) {
823 $used = true;
824 }
825 }
826
827 if ( $used ) {
828 $deps[] = new FileDependency( $fileName );
829 }
830 }
831
832 # Merge core data into extension data
833 foreach ( $coreData as $key => $item ) {
834 $this->mergeItem( $key, $allData[$key], $item );
835 }
836 wfProfileOut( __METHOD__ . '-extensions' );
837
838 # Add cache dependencies for any referenced globals
839 $deps['wgExtensionMessagesFiles'] = new GlobalDependency( 'wgExtensionMessagesFiles' );
840 $deps['version'] = new ConstantDependency( 'MW_LC_VERSION' );
841
842 # Add dependencies to the cache entry
843 $allData['deps'] = $deps;
844
845 # Replace spaces with underscores in namespace names
846 $allData['namespaceNames'] = str_replace( ' ', '_', $allData['namespaceNames'] );
847
848 # And do the same for special page aliases. $page is an array.
849 foreach ( $allData['specialPageAliases'] as &$page ) {
850 $page = str_replace( ' ', '_', $page );
851 }
852 # Decouple the reference to prevent accidental damage
853 unset( $page );
854
855 # If there were no plural rules, return an empty array
856 if ( $allData['pluralRules'] === null ) {
857 $allData['pluralRules'] = array();
858 }
859 if ( $allData['compiledPluralRules'] === null ) {
860 $allData['compiledPluralRules'] = array();
861 }
862 # If there were no plural rule types, return an empty array
863 if ( $allData['pluralRuleTypes'] === null ) {
864 $allData['pluralRuleTypes'] = array();
865 }
866
867 # Set the list keys
868 $allData['list'] = array();
869 foreach ( self::$splitKeys as $key ) {
870 $allData['list'][$key] = array_keys( $allData[$key] );
871 }
872 # Run hooks
873 wfRunHooks( 'LocalisationCacheRecache', array( $this, $code, &$allData ) );
874
875 if ( is_null( $allData['namespaceNames'] ) ) {
876 wfProfileOut( __METHOD__ );
877 throw new MWException( __METHOD__ . ': Localisation data failed sanity check! ' .
878 'Check that your languages/messages/MessagesEn.php file is intact.' );
879 }
880
881 # Set the preload key
882 $allData['preload'] = $this->buildPreload( $allData );
883
884 # Save to the process cache and register the items loaded
885 $this->data[$code] = $allData;
886 foreach ( $allData as $key => $item ) {
887 $this->loadedItems[$code][$key] = true;
888 }
889
890 # Save to the persistent cache
891 wfProfileIn( __METHOD__ . '-write' );
892 $this->store->startWrite( $code );
893 foreach ( $allData as $key => $value ) {
894 if ( in_array( $key, self::$splitKeys ) ) {
895 foreach ( $value as $subkey => $subvalue ) {
896 $this->store->set( "$key:$subkey", $subvalue );
897 }
898 } else {
899 $this->store->set( $key, $value );
900 }
901 }
902 $this->store->finishWrite();
903 wfProfileOut( __METHOD__ . '-write' );
904
905 # Clear out the MessageBlobStore
906 # HACK: If using a null (i.e. disabled) storage backend, we
907 # can't write to the MessageBlobStore either
908 if ( !$this->store instanceof LCStoreNull ) {
909 MessageBlobStore::clear();
910 }
911
912 wfProfileOut( __METHOD__ );
913 }
914
915 /**
916 * Build the preload item from the given pre-cache data.
917 *
918 * The preload item will be loaded automatically, improving performance
919 * for the commonly-requested items it contains.
920 * @param $data
921 * @return array
922 */
923 protected function buildPreload( $data ) {
924 $preload = array( 'messages' => array() );
925 foreach ( self::$preloadedKeys as $key ) {
926 $preload[$key] = $data[$key];
927 }
928
929 foreach ( $data['preloadedMessages'] as $subkey ) {
930 if ( isset( $data['messages'][$subkey] ) ) {
931 $subitem = $data['messages'][$subkey];
932 } else {
933 $subitem = null;
934 }
935 $preload['messages'][$subkey] = $subitem;
936 }
937
938 return $preload;
939 }
940
941 /**
942 * Unload the data for a given language from the object cache.
943 * Reduces memory usage.
944 * @param $code
945 */
946 public function unload( $code ) {
947 unset( $this->data[$code] );
948 unset( $this->loadedItems[$code] );
949 unset( $this->loadedSubitems[$code] );
950 unset( $this->initialisedLangs[$code] );
951 unset( $this->shallowFallbacks[$code] );
952
953 foreach ( $this->shallowFallbacks as $shallowCode => $fbCode ) {
954 if ( $fbCode === $code ) {
955 $this->unload( $shallowCode );
956 }
957 }
958 }
959
960 /**
961 * Unload all data
962 */
963 public function unloadAll() {
964 foreach ( $this->initialisedLangs as $lang => $unused ) {
965 $this->unload( $lang );
966 }
967 }
968
969 /**
970 * Disable the storage backend
971 */
972 public function disableBackend() {
973 $this->store = new LCStoreNull;
974 $this->manualRecache = false;
975 }
976 }
977
978 /**
979 * Interface for the persistence layer of LocalisationCache.
980 *
981 * The persistence layer is two-level hierarchical cache. The first level
982 * is the language, the second level is the item or subitem.
983 *
984 * Since the data for a whole language is rebuilt in one operation, it needs
985 * to have a fast and atomic method for deleting or replacing all of the
986 * current data for a given language. The interface reflects this bulk update
987 * operation. Callers writing to the cache must first call startWrite(), then
988 * will call set() a couple of thousand times, then will call finishWrite()
989 * to commit the operation. When finishWrite() is called, the cache is
990 * expected to delete all data previously stored for that language.
991 *
992 * The values stored are PHP variables suitable for serialize(). Implementations
993 * of LCStore are responsible for serializing and unserializing.
994 */
995 interface LCStore {
996 /**
997 * Get a value.
998 * @param string $code Language code
999 * @param string $key Cache key
1000 */
1001 function get( $code, $key );
1002
1003 /**
1004 * Start a write transaction.
1005 * @param string $code Language code
1006 */
1007 function startWrite( $code );
1008
1009 /**
1010 * Finish a write transaction.
1011 */
1012 function finishWrite();
1013
1014 /**
1015 * Set a key to a given value. startWrite() must be called before this
1016 * is called, and finishWrite() must be called afterwards.
1017 * @param string $key
1018 * @param mixed $value
1019 */
1020 function set( $key, $value );
1021 }
1022
1023 /**
1024 * LCStore implementation which uses PHP accelerator to store data.
1025 * This will work if one of XCache, WinCache or APC cacher is configured.
1026 * (See ObjectCache.php)
1027 */
1028 class LCStoreAccel implements LCStore {
1029 private $currentLang;
1030 private $keys;
1031
1032 public function __construct() {
1033 $this->cache = wfGetCache( CACHE_ACCEL );
1034 }
1035
1036 public function get( $code, $key ) {
1037 $k = wfMemcKey( 'l10n', $code, 'k', $key );
1038 $r = $this->cache->get( $k );
1039
1040 return $r === false ? null : $r;
1041 }
1042
1043 public function startWrite( $code ) {
1044 $k = wfMemcKey( 'l10n', $code, 'l' );
1045 $keys = $this->cache->get( $k );
1046 if ( $keys ) {
1047 foreach ( $keys as $k ) {
1048 $this->cache->delete( $k );
1049 }
1050 }
1051 $this->currentLang = $code;
1052 $this->keys = array();
1053 }
1054
1055 public function finishWrite() {
1056 if ( $this->currentLang ) {
1057 $k = wfMemcKey( 'l10n', $this->currentLang, 'l' );
1058 $this->cache->set( $k, array_keys( $this->keys ) );
1059 }
1060 $this->currentLang = null;
1061 $this->keys = array();
1062 }
1063
1064 public function set( $key, $value ) {
1065 if ( $this->currentLang ) {
1066 $k = wfMemcKey( 'l10n', $this->currentLang, 'k', $key );
1067 $this->keys[$k] = true;
1068 $this->cache->set( $k, $value );
1069 }
1070 }
1071 }
1072
1073 /**
1074 * LCStore implementation which uses the standard DB functions to store data.
1075 * This will work on any MediaWiki installation.
1076 */
1077 class LCStoreDB implements LCStore {
1078 private $currentLang;
1079 private $writesDone = false;
1080
1081 /**
1082 * @var DatabaseBase
1083 */
1084 private $dbw;
1085 private $batch;
1086 private $readOnly = false;
1087
1088 public function get( $code, $key ) {
1089 if ( $this->writesDone ) {
1090 $db = wfGetDB( DB_MASTER );
1091 } else {
1092 $db = wfGetDB( DB_SLAVE );
1093 }
1094 $row = $db->selectRow( 'l10n_cache', array( 'lc_value' ),
1095 array( 'lc_lang' => $code, 'lc_key' => $key ), __METHOD__ );
1096 if ( $row ) {
1097 return unserialize( $row->lc_value );
1098 } else {
1099 return null;
1100 }
1101 }
1102
1103 public function startWrite( $code ) {
1104 if ( $this->readOnly ) {
1105 return;
1106 }
1107
1108 if ( !$code ) {
1109 throw new MWException( __METHOD__ . ": Invalid language \"$code\"" );
1110 }
1111
1112 $this->dbw = wfGetDB( DB_MASTER );
1113 try {
1114 $this->dbw->begin( __METHOD__ );
1115 $this->dbw->delete( 'l10n_cache', array( 'lc_lang' => $code ), __METHOD__ );
1116 } catch ( DBQueryError $e ) {
1117 if ( $this->dbw->wasReadOnlyError() ) {
1118 $this->readOnly = true;
1119 $this->dbw->rollback( __METHOD__ );
1120
1121 return;
1122 } else {
1123 throw $e;
1124 }
1125 }
1126
1127 $this->currentLang = $code;
1128 $this->batch = array();
1129 }
1130
1131 public function finishWrite() {
1132 if ( $this->readOnly ) {
1133 return;
1134 }
1135
1136 if ( $this->batch ) {
1137 $this->dbw->insert( 'l10n_cache', $this->batch, __METHOD__ );
1138 }
1139
1140 $this->dbw->commit( __METHOD__ );
1141 $this->currentLang = null;
1142 $this->dbw = null;
1143 $this->batch = array();
1144 $this->writesDone = true;
1145 }
1146
1147 public function set( $key, $value ) {
1148 if ( $this->readOnly ) {
1149 return;
1150 }
1151
1152 if ( is_null( $this->currentLang ) ) {
1153 throw new MWException( __CLASS__ . ': must call startWrite() before calling set()' );
1154 }
1155
1156 $this->batch[] = array(
1157 'lc_lang' => $this->currentLang,
1158 'lc_key' => $key,
1159 'lc_value' => serialize( $value ) );
1160
1161 if ( count( $this->batch ) >= 100 ) {
1162 $this->dbw->insert( 'l10n_cache', $this->batch, __METHOD__ );
1163 $this->batch = array();
1164 }
1165 }
1166 }
1167
1168 /**
1169 * LCStore implementation which stores data as a collection of CDB files in the
1170 * directory given by $wgCacheDirectory. If $wgCacheDirectory is not set, this
1171 * will throw an exception.
1172 *
1173 * Profiling indicates that on Linux, this implementation outperforms MySQL if
1174 * the directory is on a local filesystem and there is ample kernel cache
1175 * space. The performance advantage is greater when the DBA extension is
1176 * available than it is with the PHP port.
1177 *
1178 * See Cdb.php and http://cr.yp.to/cdb.html
1179 */
1180 class LCStoreCDB implements LCStore {
1181 private $readers;
1182 private $writer;
1183 private $currentLang;
1184 private $directory;
1185
1186 function __construct( $conf = array() ) {
1187 global $wgCacheDirectory;
1188
1189 if ( isset( $conf['directory'] ) ) {
1190 $this->directory = $conf['directory'];
1191 } else {
1192 $this->directory = $wgCacheDirectory;
1193 }
1194 }
1195
1196 public function get( $code, $key ) {
1197 if ( !isset( $this->readers[$code] ) ) {
1198 $fileName = $this->getFileName( $code );
1199
1200 $this->readers[$code] = false;
1201 if ( file_exists( $fileName ) ) {
1202 try {
1203 $this->readers[$code] = CdbReader::open( $fileName );
1204 } catch ( CdbException $e ) {
1205 wfDebug( __METHOD__ . ": unable to open cdb file for reading" );
1206 }
1207 }
1208 }
1209
1210 if ( !$this->readers[$code] ) {
1211 return null;
1212 } else {
1213 $value = false;
1214 try {
1215 $value = $this->readers[$code]->get( $key );
1216 } catch ( CdbException $e ) {
1217 wfDebug( __METHOD__ . ": CdbException caught, error message was "
1218 . $e->getMessage() );
1219 }
1220 if ( $value === false ) {
1221 return null;
1222 }
1223
1224 return unserialize( $value );
1225 }
1226 }
1227
1228 public function startWrite( $code ) {
1229 if ( !file_exists( $this->directory ) ) {
1230 if ( !wfMkdirParents( $this->directory, null, __METHOD__ ) ) {
1231 throw new MWException( "Unable to create the localisation store " .
1232 "directory \"{$this->directory}\"" );
1233 }
1234 }
1235
1236 // Close reader to stop permission errors on write
1237 if ( !empty( $this->readers[$code] ) ) {
1238 $this->readers[$code]->close();
1239 }
1240
1241 try {
1242 $this->writer = CdbWriter::open( $this->getFileName( $code ) );
1243 } catch ( CdbException $e ) {
1244 throw new MWException( $e->getMessage() );
1245 }
1246 $this->currentLang = $code;
1247 }
1248
1249 public function finishWrite() {
1250 // Close the writer
1251 try {
1252 $this->writer->close();
1253 } catch ( CdbException $e ) {
1254 throw new MWException( $e->getMessage() );
1255 }
1256 $this->writer = null;
1257 unset( $this->readers[$this->currentLang] );
1258 $this->currentLang = null;
1259 }
1260
1261 public function set( $key, $value ) {
1262 if ( is_null( $this->writer ) ) {
1263 throw new MWException( __CLASS__ . ': must call startWrite() before calling set()' );
1264 }
1265 try {
1266 $this->writer->set( $key, serialize( $value ) );
1267 } catch ( CdbException $e ) {
1268 throw new MWException( $e->getMessage() );
1269 }
1270 }
1271
1272 protected function getFileName( $code ) {
1273 if ( strval( $code ) === '' || strpos( $code, '/' ) !== false ) {
1274 throw new MWException( __METHOD__ . ": Invalid language \"$code\"" );
1275 }
1276
1277 return "{$this->directory}/l10n_cache-$code.cdb";
1278 }
1279 }
1280
1281 /**
1282 * Null store backend, used to avoid DB errors during install
1283 */
1284 class LCStoreNull implements LCStore {
1285 public function get( $code, $key ) {
1286 return null;
1287 }
1288
1289 public function startWrite( $code ) {
1290 }
1291
1292 public function finishWrite() {
1293 }
1294
1295 public function set( $key, $value ) {
1296 }
1297 }
1298
1299 /**
1300 * A localisation cache optimised for loading large amounts of data for many
1301 * languages. Used by rebuildLocalisationCache.php.
1302 */
1303 class LocalisationCacheBulkLoad extends LocalisationCache {
1304 /**
1305 * A cache of the contents of data files.
1306 * Core files are serialized to avoid using ~1GB of RAM during a recache.
1307 */
1308 private $fileCache = array();
1309
1310 /**
1311 * Most recently used languages. Uses the linked-list aspect of PHP hashtables
1312 * to keep the most recently used language codes at the end of the array, and
1313 * the language codes that are ready to be deleted at the beginning.
1314 */
1315 private $mruLangs = array();
1316
1317 /**
1318 * Maximum number of languages that may be loaded into $this->data
1319 */
1320 private $maxLoadedLangs = 10;
1321
1322 /**
1323 * @param $fileName
1324 * @param $fileType
1325 * @return array|mixed
1326 */
1327 protected function readPHPFile( $fileName, $fileType ) {
1328 $serialize = $fileType === 'core';
1329 if ( !isset( $this->fileCache[$fileName][$fileType] ) ) {
1330 $data = parent::readPHPFile( $fileName, $fileType );
1331
1332 if ( $serialize ) {
1333 $encData = serialize( $data );
1334 } else {
1335 $encData = $data;
1336 }
1337
1338 $this->fileCache[$fileName][$fileType] = $encData;
1339
1340 return $data;
1341 } elseif ( $serialize ) {
1342 return unserialize( $this->fileCache[$fileName][$fileType] );
1343 } else {
1344 return $this->fileCache[$fileName][$fileType];
1345 }
1346 }
1347
1348 /**
1349 * @param $code
1350 * @param $key
1351 * @return mixed
1352 */
1353 public function getItem( $code, $key ) {
1354 unset( $this->mruLangs[$code] );
1355 $this->mruLangs[$code] = true;
1356
1357 return parent::getItem( $code, $key );
1358 }
1359
1360 /**
1361 * @param $code
1362 * @param $key
1363 * @param $subkey
1364 * @return
1365 */
1366 public function getSubitem( $code, $key, $subkey ) {
1367 unset( $this->mruLangs[$code] );
1368 $this->mruLangs[$code] = true;
1369
1370 return parent::getSubitem( $code, $key, $subkey );
1371 }
1372
1373 /**
1374 * @param $code
1375 */
1376 public function recache( $code ) {
1377 parent::recache( $code );
1378 unset( $this->mruLangs[$code] );
1379 $this->mruLangs[$code] = true;
1380 $this->trimCache();
1381 }
1382
1383 /**
1384 * @param $code
1385 */
1386 public function unload( $code ) {
1387 unset( $this->mruLangs[$code] );
1388 parent::unload( $code );
1389 }
1390
1391 /**
1392 * Unload cached languages until there are less than $this->maxLoadedLangs
1393 */
1394 protected function trimCache() {
1395 while ( count( $this->data ) > $this->maxLoadedLangs && count( $this->mruLangs ) ) {
1396 reset( $this->mruLangs );
1397 $code = key( $this->mruLangs );
1398 wfDebug( __METHOD__ . ": unloading $code\n" );
1399 $this->unload( $code );
1400 }
1401 }
1402 }