Merge "Use separate PoolCounter config for expensive thumbnails"
[mediawiki.git] / includes / cache / LocalisationCache.php
blob1153fd29620d0a20709bfbf32d630c5ce81f7d42
1 <?php
2 /**
3 * Cache of the contents of localisation files.
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
20 * @file
23 define( 'MW_LC_VERSION', 2 );
25 /**
26 * Class for caching the contents of localisation files, Messages*.php
27 * and *.i18n.php.
29 * An instance of this class is available using Language::getLocalisationCache().
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.
37 class LocalisationCache {
38 /** Configuration associative array */
39 private $conf;
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.
46 private $manualRecache = false;
48 /**
49 * True to treat all files as expired until they are regenerated by this object.
51 private $forceRecache = false;
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.
59 protected $data = array();
61 /**
62 * The persistent store object. An instance of LCStore.
64 * @var LCStore
66 private $store;
68 /**
69 * A 2-d associative array, code/key, where presence indicates that the item
70 * is loaded. Value arbitrary.
72 * For split items, if set, this indicates that all of the subitems have been
73 * loaded.
75 private $loadedItems = array();
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.
81 private $loadedSubitems = array();
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.
88 private $initialisedLangs = array();
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.
95 private $shallowFallbacks = array();
97 /**
98 * An array where the keys are codes that have been recached by this instance.
100 private $recachedLangs = array();
103 * All item keys
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',
117 * Keys for items which consist of associative arrays, which may be merged
118 * by a fallback sequence.
120 static public $mergeableMapKeys = array( 'messages', 'namespaceNames',
121 'dateFormats', 'imageFiles', 'preloadedMessages'
125 * Keys for items which are a numbered array.
127 static public $mergeableListKeys = array( 'extraUserToggles' );
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.
133 static public $mergeableAliasListKeys = array( 'specialPageAliases' );
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.
140 static public $optionalMergeKeys = array( 'bookstoreList' );
143 * Keys for items that are formatted like $magicWords
145 static public $magicWordKeys = array( 'magicWords' );
148 * Keys for items where the subitems are stored in the backend separately.
150 static public $splitKeys = array( 'messages' );
153 * Keys which are loaded automatically by initLanguage()
155 static public $preloadedKeys = array( 'dateFormats', 'namespaceNames' );
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.
161 private $pluralRules = null;
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}}.
175 private $pluralRuleTypes = null;
177 private $mergeableKeys = null;
180 * Constructor.
181 * For constructor parameters, see the documentation in DefaultSettings.php
182 * for $wgLocalisationCacheConf.
184 * @param array $conf
185 * @throws MWException
187 function __construct( $conf ) {
188 global $wgCacheDirectory;
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.' );
215 wfDebugLog( 'caches', get_class( $this ) . ": using store $storeClass" );
216 if ( !empty( $conf['storeDirectory'] ) ) {
217 $storeConf['directory'] = $conf['storeDirectory'];
220 $this->store = new $storeClass( $storeConf );
221 foreach ( array( 'manualRecache', 'forceRecache' ) as $var ) {
222 if ( isset( $conf[$var] ) ) {
223 $this->$var = $conf[$var];
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 string $key
232 * @return bool
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 ) );
245 return isset( $this->mergeableKeys[$key] );
249 * Get a cache item.
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 string $code
254 * @param string $key
255 * @return mixed
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' );
264 if ( $key === 'fallback' && isset( $this->shallowFallbacks[$code] ) ) {
265 return $this->shallowFallbacks[$code];
268 return $this->data[$code][$key];
272 * Get a subitem, for instance a single message for a given language.
273 * @param string $code
274 * @param string $key
275 * @param string $subkey
276 * @return mixed|null
278 public function getSubitem( $code, $key, $subkey ) {
279 if ( !isset( $this->loadedSubitems[$code][$key][$subkey] ) &&
280 !isset( $this->loadedItems[$code][$key] )
282 wfProfileIn( __METHOD__ . '-load' );
283 $this->loadSubitem( $code, $key, $subkey );
284 wfProfileOut( __METHOD__ . '-load' );
287 if ( isset( $this->data[$code][$key][$subkey] ) ) {
288 return $this->data[$code][$key][$subkey];
289 } else {
290 return null;
295 * Get the list of subitem keys for a given item.
297 * This is faster than array_keys($lc->getItem(...)) for the items listed in
298 * self::$splitKeys.
300 * Will return null if the item is not found, or false if the item is not an
301 * array.
302 * @param string $code
303 * @param string $key
304 * @return bool|null|string
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;
320 * Load an item into the cache.
321 * @param string $code
322 * @param string $key
324 protected function loadItem( $code, $key ) {
325 if ( !isset( $this->initialisedLangs[$code] ) ) {
326 $this->initLanguage( $code );
329 // Check to see if initLanguage() loaded it for us
330 if ( isset( $this->loadedItems[$code][$key] ) ) {
331 return;
334 if ( isset( $this->shallowFallbacks[$code] ) ) {
335 $this->loadItem( $this->shallowFallbacks[$code], $key );
337 return;
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;
346 $this->data[$code][$key][$subkey] = $this->getSubitem( $code, $key, $subkey );
348 } else {
349 $this->data[$code][$key] = $this->store->get( $code, $key );
352 $this->loadedItems[$code][$key] = true;
356 * Load a subitem into the cache
357 * @param string $code
358 * @param string $key
359 * @param string $subkey
361 protected function loadSubitem( $code, $key, $subkey ) {
362 if ( !in_array( $key, self::$splitKeys ) ) {
363 $this->loadItem( $code, $key );
365 return;
368 if ( !isset( $this->initialisedLangs[$code] ) ) {
369 $this->initLanguage( $code );
372 // Check to see if initLanguage() loaded it for us
373 if ( isset( $this->loadedItems[$code][$key] ) ||
374 isset( $this->loadedSubitems[$code][$key][$subkey] )
376 return;
379 if ( isset( $this->shallowFallbacks[$code] ) ) {
380 $this->loadSubitem( $this->shallowFallbacks[$code], $key, $subkey );
382 return;
385 $value = $this->store->get( $code, "$key:$subkey" );
386 $this->data[$code][$key][$subkey] = $value;
387 $this->loadedSubitems[$code][$key][$subkey] = true;
391 * Returns true if the cache identified by $code is missing or expired.
393 * @param string $code
395 * @return bool
397 public function isExpired( $code ) {
398 if ( $this->forceRecache && !isset( $this->recachedLangs[$code] ) ) {
399 wfDebug( __METHOD__ . "($code): forced reload\n" );
401 return true;
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" );
411 return true;
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" );
423 return true;
427 return false;
431 * Initialise a language in this object. Rebuild the cache if necessary.
432 * @param string $code
433 * @throws MWException
435 protected function initLanguage( $code ) {
436 if ( isset( $this->initialisedLangs[$code] ) ) {
437 return;
440 $this->initialisedLangs[$code] = true;
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( $code, 'en' );
446 return;
449 # Recache the data if necessary
450 if ( !$this->manualRecache && $this->isExpired( $code ) ) {
451 if ( Language::isSupportedLanguage( $code ) ) {
452 $this->recache( $code );
453 } elseif ( $code === 'en' ) {
454 throw new MWException( 'MessagesEn.php is missing.' );
455 } else {
456 $this->initShallowFallback( $code, 'en' );
459 return;
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.' );
471 $this->initShallowFallback( $code, 'en' );
473 return;
474 } else {
475 throw new MWException( 'Invalid or missing localisation cache.' );
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;
484 } else {
485 $this->loadedItems[$code][$key] = true;
491 * Create a fallback from one language to another, without creating a
492 * complete persistent cache.
493 * @param string $primaryCode
494 * @param string $fallbackCode
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;
504 * Read a PHP file containing localisation data.
505 * @param string $_fileName
506 * @param string $_fileType
507 * @throws MWException
508 * @return array
510 protected function readPHPFile( $_fileName, $_fileType ) {
511 wfProfileIn( __METHOD__ );
512 // Disable APC caching
513 wfSuppressWarnings();
514 $_apcEnabled = ini_set( 'apc.cache_by_default', '0' );
515 wfRestoreWarnings();
517 include $_fileName;
519 wfSuppressWarnings();
520 ini_set( 'apc.cache_by_default', $_apcEnabled );
521 wfRestoreWarnings();
523 if ( $_fileType == 'core' || $_fileType == 'extension' ) {
524 $data = compact( self::$allKeys );
525 } elseif ( $_fileType == 'aliases' ) {
526 $data = compact( 'aliases' );
527 } else {
528 wfProfileOut( __METHOD__ );
529 throw new MWException( __METHOD__ . ": Invalid file type: $_fileType" );
531 wfProfileOut( __METHOD__ );
533 return $data;
537 * Read a JSON file containing localisation messages.
538 * @param string $fileName Name of file to read
539 * @throws MWException if there is a syntax error in the JSON file
540 * @return array Array with a 'messages' key, or empty array if the file doesn't exist
542 public function readJSONFile( $fileName ) {
543 wfProfileIn( __METHOD__ );
545 if ( !is_readable( $fileName ) ) {
546 wfProfileOut( __METHOD__ );
548 return array();
551 $json = file_get_contents( $fileName );
552 if ( $json === false ) {
553 wfProfileOut( __METHOD__ );
555 return array();
558 $data = FormatJson::decode( $json, true );
559 if ( $data === null ) {
560 wfProfileOut( __METHOD__ );
562 throw new MWException( __METHOD__ . ": Invalid JSON file: $fileName" );
565 // Remove keys starting with '@', they're reserved for metadata and non-message data
566 foreach ( $data as $key => $unused ) {
567 if ( $key === '' || $key[0] === '@' ) {
568 unset( $data[$key] );
572 wfProfileOut( __METHOD__ );
574 // The JSON format only supports messages, none of the other variables, so wrap the data
575 return array( 'messages' => $data );
579 * Get the compiled plural rules for a given language from the XML files.
580 * @since 1.20
581 * @param string $code
582 * @return array|null
584 public function getCompiledPluralRules( $code ) {
585 $rules = $this->getPluralRules( $code );
586 if ( $rules === null ) {
587 return null;
589 try {
590 $compiledRules = CLDRPluralRuleEvaluator::compile( $rules );
591 } catch ( CLDRPluralRuleError $e ) {
592 wfDebugLog( 'l10n', $e->getMessage() );
594 return array();
597 return $compiledRules;
601 * Get the plural rules for a given language from the XML files.
602 * Cached.
603 * @since 1.20
604 * @param string $code
605 * @return array|null
607 public function getPluralRules( $code ) {
608 if ( $this->pluralRules === null ) {
609 $this->loadPluralFiles();
611 if ( !isset( $this->pluralRules[$code] ) ) {
612 return null;
613 } else {
614 return $this->pluralRules[$code];
619 * Get the plural rule types for a given language from the XML files.
620 * Cached.
621 * @since 1.22
622 * @param string $code
623 * @return array|null
625 public function getPluralRuleTypes( $code ) {
626 if ( $this->pluralRuleTypes === null ) {
627 $this->loadPluralFiles();
629 if ( !isset( $this->pluralRuleTypes[$code] ) ) {
630 return null;
631 } else {
632 return $this->pluralRuleTypes[$code];
637 * Load the plural XML files.
639 protected function loadPluralFiles() {
640 global $IP;
641 $cldrPlural = "$IP/languages/data/plurals.xml";
642 $mwPlural = "$IP/languages/data/plurals-mediawiki.xml";
643 // Load CLDR plural rules
644 $this->loadPluralFile( $cldrPlural );
645 if ( file_exists( $mwPlural ) ) {
646 // Override or extend
647 $this->loadPluralFile( $mwPlural );
652 * Load a plural XML file with the given filename, compile the relevant
653 * rules, and save the compiled rules in a process-local cache.
655 * @param string $fileName
657 protected function loadPluralFile( $fileName ) {
658 $doc = new DOMDocument;
659 $doc->load( $fileName );
660 $rulesets = $doc->getElementsByTagName( "pluralRules" );
661 foreach ( $rulesets as $ruleset ) {
662 $codes = $ruleset->getAttribute( 'locales' );
663 $rules = array();
664 $ruleTypes = array();
665 $ruleElements = $ruleset->getElementsByTagName( "pluralRule" );
666 foreach ( $ruleElements as $elt ) {
667 $ruleType = $elt->getAttribute( 'count' );
668 if ( $ruleType === 'other' ) {
669 // Don't record "other" rules, which have an empty condition
670 continue;
672 $rules[] = $elt->nodeValue;
673 $ruleTypes[] = $ruleType;
675 foreach ( explode( ' ', $codes ) as $code ) {
676 $this->pluralRules[$code] = $rules;
677 $this->pluralRuleTypes[$code] = $ruleTypes;
683 * Read the data from the source files for a given language, and register
684 * the relevant dependencies in the $deps array. If the localisation
685 * exists, the data array is returned, otherwise false is returned.
687 * @param string $code
688 * @param array $deps
690 protected function readSourceFilesAndRegisterDeps( $code, &$deps ) {
691 global $IP;
692 wfProfileIn( __METHOD__ );
694 // This reads in the PHP i18n file with non-messages l10n data
695 $fileName = Language::getMessagesFileName( $code );
696 if ( !file_exists( $fileName ) ) {
697 $data = array();
698 } else {
699 $deps[] = new FileDependency( $fileName );
700 $data = $this->readPHPFile( $fileName, 'core' );
703 # Load CLDR plural rules for JavaScript
704 $data['pluralRules'] = $this->getPluralRules( $code );
705 # And for PHP
706 $data['compiledPluralRules'] = $this->getCompiledPluralRules( $code );
707 # Load plural rule types
708 $data['pluralRuleTypes'] = $this->getPluralRuleTypes( $code );
710 $deps['plurals'] = new FileDependency( "$IP/languages/data/plurals.xml" );
711 $deps['plurals-mw'] = new FileDependency( "$IP/languages/data/plurals-mediawiki.xml" );
713 wfProfileOut( __METHOD__ );
715 return $data;
719 * Merge two localisation values, a primary and a fallback, overwriting the
720 * primary value in place.
721 * @param string $key
722 * @param mixed $value
723 * @param mixed $fallbackValue
725 protected function mergeItem( $key, &$value, $fallbackValue ) {
726 if ( !is_null( $value ) ) {
727 if ( !is_null( $fallbackValue ) ) {
728 if ( in_array( $key, self::$mergeableMapKeys ) ) {
729 $value = $value + $fallbackValue;
730 } elseif ( in_array( $key, self::$mergeableListKeys ) ) {
731 $value = array_unique( array_merge( $fallbackValue, $value ) );
732 } elseif ( in_array( $key, self::$mergeableAliasListKeys ) ) {
733 $value = array_merge_recursive( $value, $fallbackValue );
734 } elseif ( in_array( $key, self::$optionalMergeKeys ) ) {
735 if ( !empty( $value['inherit'] ) ) {
736 $value = array_merge( $fallbackValue, $value );
739 if ( isset( $value['inherit'] ) ) {
740 unset( $value['inherit'] );
742 } elseif ( in_array( $key, self::$magicWordKeys ) ) {
743 $this->mergeMagicWords( $value, $fallbackValue );
746 } else {
747 $value = $fallbackValue;
752 * @param mixed $value
753 * @param mixed $fallbackValue
755 protected function mergeMagicWords( &$value, $fallbackValue ) {
756 foreach ( $fallbackValue as $magicName => $fallbackInfo ) {
757 if ( !isset( $value[$magicName] ) ) {
758 $value[$magicName] = $fallbackInfo;
759 } else {
760 $oldSynonyms = array_slice( $fallbackInfo, 1 );
761 $newSynonyms = array_slice( $value[$magicName], 1 );
762 $synonyms = array_values( array_unique( array_merge(
763 $newSynonyms, $oldSynonyms ) ) );
764 $value[$magicName] = array_merge( array( $fallbackInfo[0] ), $synonyms );
770 * Given an array mapping language code to localisation value, such as is
771 * found in extension *.i18n.php files, iterate through a fallback sequence
772 * to merge the given data with an existing primary value.
774 * Returns true if any data from the extension array was used, false
775 * otherwise.
776 * @param string $codeSequence
777 * @param string $key
778 * @param mixed $value
779 * @param mixed $fallbackValue
780 * @return bool
782 protected function mergeExtensionItem( $codeSequence, $key, &$value, $fallbackValue ) {
783 $used = false;
784 foreach ( $codeSequence as $code ) {
785 if ( isset( $fallbackValue[$code] ) ) {
786 $this->mergeItem( $key, $value, $fallbackValue[$code] );
787 $used = true;
791 return $used;
795 * Load localisation data for a given language for both core and extensions
796 * and save it to the persistent cache store and the process cache
797 * @param string $code
798 * @throws MWException
800 public function recache( $code ) {
801 global $wgExtensionMessagesFiles, $wgMessagesDirs;
802 wfProfileIn( __METHOD__ );
804 if ( !$code ) {
805 wfProfileOut( __METHOD__ );
806 throw new MWException( "Invalid language code requested" );
808 $this->recachedLangs[$code] = true;
810 # Initial values
811 $initialData = array_combine(
812 self::$allKeys,
813 array_fill( 0, count( self::$allKeys ), null ) );
814 $coreData = $initialData;
815 $deps = array();
817 # Load the primary localisation from the source file
818 $data = $this->readSourceFilesAndRegisterDeps( $code, $deps );
819 if ( $data === false ) {
820 wfDebug( __METHOD__ . ": no localisation file for $code, using fallback to en\n" );
821 $coreData['fallback'] = 'en';
822 } else {
823 wfDebug( __METHOD__ . ": got localisation for $code from source\n" );
825 # Merge primary localisation
826 foreach ( $data as $key => $value ) {
827 $this->mergeItem( $key, $coreData[$key], $value );
831 # Fill in the fallback if it's not there already
832 if ( is_null( $coreData['fallback'] ) ) {
833 $coreData['fallback'] = $code === 'en' ? false : 'en';
835 if ( $coreData['fallback'] === false ) {
836 $coreData['fallbackSequence'] = array();
837 } else {
838 $coreData['fallbackSequence'] = array_map( 'trim', explode( ',', $coreData['fallback'] ) );
839 $len = count( $coreData['fallbackSequence'] );
841 # Ensure that the sequence ends at en
842 if ( $coreData['fallbackSequence'][$len - 1] !== 'en' ) {
843 $coreData['fallbackSequence'][] = 'en';
846 # Load the fallback localisation item by item and merge it
847 foreach ( $coreData['fallbackSequence'] as $fbCode ) {
848 # Load the secondary localisation from the source file to
849 # avoid infinite cycles on cyclic fallbacks
850 $fbData = $this->readSourceFilesAndRegisterDeps( $fbCode, $deps );
851 if ( $fbData === false ) {
852 continue;
855 foreach ( self::$allKeys as $key ) {
856 if ( !isset( $fbData[$key] ) ) {
857 continue;
860 if ( is_null( $coreData[$key] ) || $this->isMergeableKey( $key ) ) {
861 $this->mergeItem( $key, $coreData[$key], $fbData[$key] );
867 $codeSequence = array_merge( array( $code ), $coreData['fallbackSequence'] );
869 # Load core messages and the extension localisations.
870 wfProfileIn( __METHOD__ . '-extensions' );
871 $allData = $initialData;
872 foreach ( $wgMessagesDirs as $dirs ) {
873 foreach ( (array)$dirs as $dir ) {
874 foreach ( $codeSequence as $csCode ) {
875 $fileName = "$dir/$csCode.json";
876 $data = $this->readJSONFile( $fileName );
878 foreach ( $data as $key => $item ) {
879 $this->mergeItem( $key, $allData[$key], $item );
882 $deps[] = new FileDependency( $fileName );
887 foreach ( $wgExtensionMessagesFiles as $extension => $fileName ) {
888 if ( isset( $wgMessagesDirs[$extension] ) ) {
889 # Already loaded the JSON files for this extension; skip the PHP shim
890 continue;
893 $data = $this->readPHPFile( $fileName, 'extension' );
894 $used = false;
896 foreach ( $data as $key => $item ) {
897 if ( $this->mergeExtensionItem( $codeSequence, $key, $allData[$key], $item ) ) {
898 $used = true;
902 if ( $used ) {
903 $deps[] = new FileDependency( $fileName );
907 # Merge core data into extension data
908 foreach ( $coreData as $key => $item ) {
909 $this->mergeItem( $key, $allData[$key], $item );
911 wfProfileOut( __METHOD__ . '-extensions' );
913 # Add cache dependencies for any referenced globals
914 $deps['wgExtensionMessagesFiles'] = new GlobalDependency( 'wgExtensionMessagesFiles' );
915 $deps['wgMessagesDirs'] = new GlobalDependency( 'wgMessagesDirs' );
916 $deps['version'] = new ConstantDependency( 'MW_LC_VERSION' );
918 # Add dependencies to the cache entry
919 $allData['deps'] = $deps;
921 # Replace spaces with underscores in namespace names
922 $allData['namespaceNames'] = str_replace( ' ', '_', $allData['namespaceNames'] );
924 # And do the same for special page aliases. $page is an array.
925 foreach ( $allData['specialPageAliases'] as &$page ) {
926 $page = str_replace( ' ', '_', $page );
928 # Decouple the reference to prevent accidental damage
929 unset( $page );
931 # If there were no plural rules, return an empty array
932 if ( $allData['pluralRules'] === null ) {
933 $allData['pluralRules'] = array();
935 if ( $allData['compiledPluralRules'] === null ) {
936 $allData['compiledPluralRules'] = array();
938 # If there were no plural rule types, return an empty array
939 if ( $allData['pluralRuleTypes'] === null ) {
940 $allData['pluralRuleTypes'] = array();
943 # Set the list keys
944 $allData['list'] = array();
945 foreach ( self::$splitKeys as $key ) {
946 $allData['list'][$key] = array_keys( $allData[$key] );
948 # Run hooks
949 $purgeBlobs = true;
950 wfRunHooks( 'LocalisationCacheRecache', array( $this, $code, &$allData, &$purgeBlobs ) );
952 if ( is_null( $allData['namespaceNames'] ) ) {
953 wfProfileOut( __METHOD__ );
954 throw new MWException( __METHOD__ . ': Localisation data failed sanity check! ' .
955 'Check that your languages/messages/MessagesEn.php file is intact.' );
958 # Set the preload key
959 $allData['preload'] = $this->buildPreload( $allData );
961 # Save to the process cache and register the items loaded
962 $this->data[$code] = $allData;
963 foreach ( $allData as $key => $item ) {
964 $this->loadedItems[$code][$key] = true;
967 # Save to the persistent cache
968 wfProfileIn( __METHOD__ . '-write' );
969 $this->store->startWrite( $code );
970 foreach ( $allData as $key => $value ) {
971 if ( in_array( $key, self::$splitKeys ) ) {
972 foreach ( $value as $subkey => $subvalue ) {
973 $this->store->set( "$key:$subkey", $subvalue );
975 } else {
976 $this->store->set( $key, $value );
979 $this->store->finishWrite();
980 wfProfileOut( __METHOD__ . '-write' );
982 # Clear out the MessageBlobStore
983 # HACK: If using a null (i.e. disabled) storage backend, we
984 # can't write to the MessageBlobStore either
985 if ( $purgeBlobs && !$this->store instanceof LCStoreNull ) {
986 MessageBlobStore::clear();
989 wfProfileOut( __METHOD__ );
993 * Build the preload item from the given pre-cache data.
995 * The preload item will be loaded automatically, improving performance
996 * for the commonly-requested items it contains.
997 * @param array $data
998 * @return array
1000 protected function buildPreload( $data ) {
1001 $preload = array( 'messages' => array() );
1002 foreach ( self::$preloadedKeys as $key ) {
1003 $preload[$key] = $data[$key];
1006 foreach ( $data['preloadedMessages'] as $subkey ) {
1007 if ( isset( $data['messages'][$subkey] ) ) {
1008 $subitem = $data['messages'][$subkey];
1009 } else {
1010 $subitem = null;
1012 $preload['messages'][$subkey] = $subitem;
1015 return $preload;
1019 * Unload the data for a given language from the object cache.
1020 * Reduces memory usage.
1021 * @param string $code
1023 public function unload( $code ) {
1024 unset( $this->data[$code] );
1025 unset( $this->loadedItems[$code] );
1026 unset( $this->loadedSubitems[$code] );
1027 unset( $this->initialisedLangs[$code] );
1028 unset( $this->shallowFallbacks[$code] );
1030 foreach ( $this->shallowFallbacks as $shallowCode => $fbCode ) {
1031 if ( $fbCode === $code ) {
1032 $this->unload( $shallowCode );
1038 * Unload all data
1040 public function unloadAll() {
1041 foreach ( $this->initialisedLangs as $lang => $unused ) {
1042 $this->unload( $lang );
1047 * Disable the storage backend
1049 public function disableBackend() {
1050 $this->store = new LCStoreNull;
1051 $this->manualRecache = false;
1056 * Interface for the persistence layer of LocalisationCache.
1058 * The persistence layer is two-level hierarchical cache. The first level
1059 * is the language, the second level is the item or subitem.
1061 * Since the data for a whole language is rebuilt in one operation, it needs
1062 * to have a fast and atomic method for deleting or replacing all of the
1063 * current data for a given language. The interface reflects this bulk update
1064 * operation. Callers writing to the cache must first call startWrite(), then
1065 * will call set() a couple of thousand times, then will call finishWrite()
1066 * to commit the operation. When finishWrite() is called, the cache is
1067 * expected to delete all data previously stored for that language.
1069 * The values stored are PHP variables suitable for serialize(). Implementations
1070 * of LCStore are responsible for serializing and unserializing.
1072 interface LCStore {
1074 * Get a value.
1075 * @param string $code Language code
1076 * @param string $key Cache key
1078 function get( $code, $key );
1081 * Start a write transaction.
1082 * @param string $code Language code
1084 function startWrite( $code );
1087 * Finish a write transaction.
1089 function finishWrite();
1092 * Set a key to a given value. startWrite() must be called before this
1093 * is called, and finishWrite() must be called afterwards.
1094 * @param string $key
1095 * @param mixed $value
1097 function set( $key, $value );
1101 * LCStore implementation which uses PHP accelerator to store data.
1102 * This will work if one of XCache, WinCache or APC cacher is configured.
1103 * (See ObjectCache.php)
1105 class LCStoreAccel implements LCStore {
1106 private $currentLang;
1107 private $keys;
1109 public function __construct() {
1110 $this->cache = wfGetCache( CACHE_ACCEL );
1113 public function get( $code, $key ) {
1114 $k = wfMemcKey( 'l10n', $code, 'k', $key );
1115 $r = $this->cache->get( $k );
1117 return $r === false ? null : $r;
1120 public function startWrite( $code ) {
1121 $k = wfMemcKey( 'l10n', $code, 'l' );
1122 $keys = $this->cache->get( $k );
1123 if ( $keys ) {
1124 foreach ( $keys as $k ) {
1125 $this->cache->delete( $k );
1128 $this->currentLang = $code;
1129 $this->keys = array();
1132 public function finishWrite() {
1133 if ( $this->currentLang ) {
1134 $k = wfMemcKey( 'l10n', $this->currentLang, 'l' );
1135 $this->cache->set( $k, array_keys( $this->keys ) );
1137 $this->currentLang = null;
1138 $this->keys = array();
1141 public function set( $key, $value ) {
1142 if ( $this->currentLang ) {
1143 $k = wfMemcKey( 'l10n', $this->currentLang, 'k', $key );
1144 $this->keys[$k] = true;
1145 $this->cache->set( $k, $value );
1151 * LCStore implementation which uses the standard DB functions to store data.
1152 * This will work on any MediaWiki installation.
1154 class LCStoreDB implements LCStore {
1155 private $currentLang;
1156 private $writesDone = false;
1159 * @var DatabaseBase
1161 private $dbw;
1162 private $batch;
1163 private $readOnly = false;
1165 public function get( $code, $key ) {
1166 if ( $this->writesDone ) {
1167 $db = wfGetDB( DB_MASTER );
1168 } else {
1169 $db = wfGetDB( DB_SLAVE );
1171 $row = $db->selectRow( 'l10n_cache', array( 'lc_value' ),
1172 array( 'lc_lang' => $code, 'lc_key' => $key ), __METHOD__ );
1173 if ( $row ) {
1174 return unserialize( $db->decodeBlob( $row->lc_value ) );
1175 } else {
1176 return null;
1180 public function startWrite( $code ) {
1181 if ( $this->readOnly ) {
1182 return;
1185 if ( !$code ) {
1186 throw new MWException( __METHOD__ . ": Invalid language \"$code\"" );
1189 $this->dbw = wfGetDB( DB_MASTER );
1190 try {
1191 $this->dbw->begin( __METHOD__ );
1192 $this->dbw->delete( 'l10n_cache', array( 'lc_lang' => $code ), __METHOD__ );
1193 } catch ( DBQueryError $e ) {
1194 if ( $this->dbw->wasReadOnlyError() ) {
1195 $this->readOnly = true;
1196 $this->dbw->rollback( __METHOD__ );
1198 return;
1199 } else {
1200 throw $e;
1204 $this->currentLang = $code;
1205 $this->batch = array();
1208 public function finishWrite() {
1209 if ( $this->readOnly ) {
1210 return;
1213 if ( $this->batch ) {
1214 $this->dbw->insert( 'l10n_cache', $this->batch, __METHOD__ );
1217 $this->dbw->commit( __METHOD__ );
1218 $this->currentLang = null;
1219 $this->dbw = null;
1220 $this->batch = array();
1221 $this->writesDone = true;
1224 public function set( $key, $value ) {
1225 if ( $this->readOnly ) {
1226 return;
1229 if ( is_null( $this->currentLang ) ) {
1230 throw new MWException( __CLASS__ . ': must call startWrite() before calling set()' );
1233 $this->batch[] = array(
1234 'lc_lang' => $this->currentLang,
1235 'lc_key' => $key,
1236 'lc_value' => $this->dbw->encodeBlob( serialize( $value ) ) );
1238 if ( count( $this->batch ) >= 100 ) {
1239 $this->dbw->insert( 'l10n_cache', $this->batch, __METHOD__ );
1240 $this->batch = array();
1246 * LCStore implementation which stores data as a collection of CDB files in the
1247 * directory given by $wgCacheDirectory. If $wgCacheDirectory is not set, this
1248 * will throw an exception.
1250 * Profiling indicates that on Linux, this implementation outperforms MySQL if
1251 * the directory is on a local filesystem and there is ample kernel cache
1252 * space. The performance advantage is greater when the DBA extension is
1253 * available than it is with the PHP port.
1255 * See Cdb.php and http://cr.yp.to/cdb.html
1257 class LCStoreCDB implements LCStore {
1258 /** @var CdbReader[] */
1259 private $readers;
1261 /** @var CdbWriter */
1262 private $writer;
1264 /** @var string Current language code */
1265 private $currentLang;
1267 /** @var bool|string Cache directory. False if not set */
1268 private $directory;
1270 function __construct( $conf = array() ) {
1271 global $wgCacheDirectory;
1273 if ( isset( $conf['directory'] ) ) {
1274 $this->directory = $conf['directory'];
1275 } else {
1276 $this->directory = $wgCacheDirectory;
1280 public function get( $code, $key ) {
1281 if ( !isset( $this->readers[$code] ) ) {
1282 $fileName = $this->getFileName( $code );
1284 $this->readers[$code] = false;
1285 if ( file_exists( $fileName ) ) {
1286 try {
1287 $this->readers[$code] = CdbReader::open( $fileName );
1288 } catch ( CdbException $e ) {
1289 wfDebug( __METHOD__ . ": unable to open cdb file for reading\n" );
1294 if ( !$this->readers[$code] ) {
1295 return null;
1296 } else {
1297 $value = false;
1298 try {
1299 $value = $this->readers[$code]->get( $key );
1300 } catch ( CdbException $e ) {
1301 wfDebug( __METHOD__ . ": CdbException caught, error message was "
1302 . $e->getMessage() . "\n" );
1304 if ( $value === false ) {
1305 return null;
1308 return unserialize( $value );
1312 public function startWrite( $code ) {
1313 if ( !file_exists( $this->directory ) ) {
1314 if ( !wfMkdirParents( $this->directory, null, __METHOD__ ) ) {
1315 throw new MWException( "Unable to create the localisation store " .
1316 "directory \"{$this->directory}\"" );
1320 // Close reader to stop permission errors on write
1321 if ( !empty( $this->readers[$code] ) ) {
1322 $this->readers[$code]->close();
1325 try {
1326 $this->writer = CdbWriter::open( $this->getFileName( $code ) );
1327 } catch ( CdbException $e ) {
1328 throw new MWException( $e->getMessage() );
1330 $this->currentLang = $code;
1333 public function finishWrite() {
1334 // Close the writer
1335 try {
1336 $this->writer->close();
1337 } catch ( CdbException $e ) {
1338 throw new MWException( $e->getMessage() );
1340 $this->writer = null;
1341 unset( $this->readers[$this->currentLang] );
1342 $this->currentLang = null;
1345 public function set( $key, $value ) {
1346 if ( is_null( $this->writer ) ) {
1347 throw new MWException( __CLASS__ . ': must call startWrite() before calling set()' );
1349 try {
1350 $this->writer->set( $key, serialize( $value ) );
1351 } catch ( CdbException $e ) {
1352 throw new MWException( $e->getMessage() );
1356 protected function getFileName( $code ) {
1357 if ( strval( $code ) === '' || strpos( $code, '/' ) !== false ) {
1358 throw new MWException( __METHOD__ . ": Invalid language \"$code\"" );
1361 return "{$this->directory}/l10n_cache-$code.cdb";
1366 * Null store backend, used to avoid DB errors during install
1368 class LCStoreNull implements LCStore {
1369 public function get( $code, $key ) {
1370 return null;
1373 public function startWrite( $code ) {
1376 public function finishWrite() {
1379 public function set( $key, $value ) {
1384 * A localisation cache optimised for loading large amounts of data for many
1385 * languages. Used by rebuildLocalisationCache.php.
1387 class LocalisationCacheBulkLoad extends LocalisationCache {
1389 * A cache of the contents of data files.
1390 * Core files are serialized to avoid using ~1GB of RAM during a recache.
1392 private $fileCache = array();
1395 * Most recently used languages. Uses the linked-list aspect of PHP hashtables
1396 * to keep the most recently used language codes at the end of the array, and
1397 * the language codes that are ready to be deleted at the beginning.
1399 private $mruLangs = array();
1402 * Maximum number of languages that may be loaded into $this->data
1404 private $maxLoadedLangs = 10;
1407 * @param string $fileName
1408 * @param string $fileType
1409 * @return array|mixed
1411 protected function readPHPFile( $fileName, $fileType ) {
1412 $serialize = $fileType === 'core';
1413 if ( !isset( $this->fileCache[$fileName][$fileType] ) ) {
1414 $data = parent::readPHPFile( $fileName, $fileType );
1416 if ( $serialize ) {
1417 $encData = serialize( $data );
1418 } else {
1419 $encData = $data;
1422 $this->fileCache[$fileName][$fileType] = $encData;
1424 return $data;
1425 } elseif ( $serialize ) {
1426 return unserialize( $this->fileCache[$fileName][$fileType] );
1427 } else {
1428 return $this->fileCache[$fileName][$fileType];
1433 * @param string $code
1434 * @param string $key
1435 * @return mixed
1437 public function getItem( $code, $key ) {
1438 unset( $this->mruLangs[$code] );
1439 $this->mruLangs[$code] = true;
1441 return parent::getItem( $code, $key );
1445 * @param string $code
1446 * @param string $key
1447 * @param string $subkey
1448 * @return mixed
1450 public function getSubitem( $code, $key, $subkey ) {
1451 unset( $this->mruLangs[$code] );
1452 $this->mruLangs[$code] = true;
1454 return parent::getSubitem( $code, $key, $subkey );
1458 * @param string $code
1460 public function recache( $code ) {
1461 parent::recache( $code );
1462 unset( $this->mruLangs[$code] );
1463 $this->mruLangs[$code] = true;
1464 $this->trimCache();
1468 * @param string $code
1470 public function unload( $code ) {
1471 unset( $this->mruLangs[$code] );
1472 parent::unload( $code );
1476 * Unload cached languages until there are less than $this->maxLoadedLangs
1478 protected function trimCache() {
1479 while ( count( $this->data ) > $this->maxLoadedLangs && count( $this->mruLangs ) ) {
1480 reset( $this->mruLangs );
1481 $code = key( $this->mruLangs );
1482 wfDebug( __METHOD__ . ": unloading $code\n" );
1483 $this->unload( $code );