Merge "Made Title cache use MapCacheLRU"
[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 * Read a JSON file containing localisation messages.
532 * @param string $fileName Name of file to read
533 * @throws MWException if there is a syntax error in the JSON file
534 * @return array with a 'messages' key, or empty array if the file doesn't exist
535 */
536 protected function readJSONFile( $fileName ) {
537 wfProfileIn( __METHOD__ );
538 if ( !is_readable( $fileName ) ) {
539 return array();
540 }
541
542 $json = file_get_contents( $fileName );
543 if ( $json === false ) {
544 return array();
545 }
546 $data = FormatJson::decode( $json, true );
547 if ( $data === null ) {
548 throw new MWException( __METHOD__ . ": Invalid JSON file: $fileName" );
549 }
550 // Remove keys starting with '@', they're reserved for metadata and non-message data
551 foreach ( $data as $key => $unused ) {
552 if ( $key === '' || $key[0] === '@' ) {
553 unset( $data[$key] );
554 }
555 }
556 // The JSON format only supports messages, none of the other variables, so wrap the data
557 return array( 'messages' => $data );
558 }
559
560 /**
561 * Get the compiled plural rules for a given language from the XML files.
562 * @since 1.20
563 */
564 public function getCompiledPluralRules( $code ) {
565 $rules = $this->getPluralRules( $code );
566 if ( $rules === null ) {
567 return null;
568 }
569 try {
570 $compiledRules = CLDRPluralRuleEvaluator::compile( $rules );
571 } catch ( CLDRPluralRuleError $e ) {
572 wfDebugLog( 'l10n', $e->getMessage() . "\n" );
573
574 return array();
575 }
576
577 return $compiledRules;
578 }
579
580 /**
581 * Get the plural rules for a given language from the XML files.
582 * Cached.
583 * @since 1.20
584 */
585 public function getPluralRules( $code ) {
586 if ( $this->pluralRules === null ) {
587 $this->loadPluralFiles();
588 }
589 if ( !isset( $this->pluralRules[$code] ) ) {
590 return null;
591 } else {
592 return $this->pluralRules[$code];
593 }
594 }
595
596 /**
597 * Get the plural rule types for a given language from the XML files.
598 * Cached.
599 * @since 1.22
600 */
601 public function getPluralRuleTypes( $code ) {
602 if ( $this->pluralRuleTypes === null ) {
603 $this->loadPluralFiles();
604 }
605 if ( !isset( $this->pluralRuleTypes[$code] ) ) {
606 return null;
607 } else {
608 return $this->pluralRuleTypes[$code];
609 }
610 }
611
612 /**
613 * Load the plural XML files.
614 */
615 protected function loadPluralFiles() {
616 global $IP;
617 $cldrPlural = "$IP/languages/data/plurals.xml";
618 $mwPlural = "$IP/languages/data/plurals-mediawiki.xml";
619 // Load CLDR plural rules
620 $this->loadPluralFile( $cldrPlural );
621 if ( file_exists( $mwPlural ) ) {
622 // Override or extend
623 $this->loadPluralFile( $mwPlural );
624 }
625 }
626
627 /**
628 * Load a plural XML file with the given filename, compile the relevant
629 * rules, and save the compiled rules in a process-local cache.
630 */
631 protected function loadPluralFile( $fileName ) {
632 $doc = new DOMDocument;
633 $doc->load( $fileName );
634 $rulesets = $doc->getElementsByTagName( "pluralRules" );
635 foreach ( $rulesets as $ruleset ) {
636 $codes = $ruleset->getAttribute( 'locales' );
637 $rules = array();
638 $ruleTypes = array();
639 $ruleElements = $ruleset->getElementsByTagName( "pluralRule" );
640 foreach ( $ruleElements as $elt ) {
641 $ruleType = $elt->getAttribute( 'count' );
642 if ( $ruleType === 'other' ) {
643 // Don't record "other" rules, which have an empty condition
644 continue;
645 }
646 $rules[] = $elt->nodeValue;
647 $ruleTypes[] = $ruleType;
648 }
649 foreach ( explode( ' ', $codes ) as $code ) {
650 $this->pluralRules[$code] = $rules;
651 $this->pluralRuleTypes[$code] = $ruleTypes;
652 }
653 }
654 }
655
656 /**
657 * Read the data from the source files for a given language, and register
658 * the relevant dependencies in the $deps array. If the localisation
659 * exists, the data array is returned, otherwise false is returned.
660 */
661 protected function readSourceFilesAndRegisterDeps( $code, &$deps ) {
662 global $IP;
663 wfProfileIn( __METHOD__ );
664
665 $fileName = Language::getMessagesFileName( $code );
666 if ( !file_exists( $fileName ) ) {
667 wfProfileOut( __METHOD__ );
668
669 return false;
670 }
671
672 $deps[] = new FileDependency( $fileName );
673 $data = $this->readPHPFile( $fileName, 'core' );
674
675 # Load CLDR plural rules for JavaScript
676 $data['pluralRules'] = $this->getPluralRules( $code );
677 # And for PHP
678 $data['compiledPluralRules'] = $this->getCompiledPluralRules( $code );
679 # Load plural rule types
680 $data['pluralRuleTypes'] = $this->getPluralRuleTypes( $code );
681
682 $deps['plurals'] = new FileDependency( "$IP/languages/data/plurals.xml" );
683 $deps['plurals-mw'] = new FileDependency( "$IP/languages/data/plurals-mediawiki.xml" );
684
685 wfProfileOut( __METHOD__ );
686
687 return $data;
688 }
689
690 /**
691 * Merge two localisation values, a primary and a fallback, overwriting the
692 * primary value in place.
693 * @param $key
694 * @param $value
695 * @param $fallbackValue
696 */
697 protected function mergeItem( $key, &$value, $fallbackValue ) {
698 if ( !is_null( $value ) ) {
699 if ( !is_null( $fallbackValue ) ) {
700 if ( in_array( $key, self::$mergeableMapKeys ) ) {
701 $value = $value + $fallbackValue;
702 } elseif ( in_array( $key, self::$mergeableListKeys ) ) {
703 $value = array_unique( array_merge( $fallbackValue, $value ) );
704 } elseif ( in_array( $key, self::$mergeableAliasListKeys ) ) {
705 $value = array_merge_recursive( $value, $fallbackValue );
706 } elseif ( in_array( $key, self::$optionalMergeKeys ) ) {
707 if ( !empty( $value['inherit'] ) ) {
708 $value = array_merge( $fallbackValue, $value );
709 }
710
711 if ( isset( $value['inherit'] ) ) {
712 unset( $value['inherit'] );
713 }
714 } elseif ( in_array( $key, self::$magicWordKeys ) ) {
715 $this->mergeMagicWords( $value, $fallbackValue );
716 }
717 }
718 } else {
719 $value = $fallbackValue;
720 }
721 }
722
723 /**
724 * @param $value
725 * @param $fallbackValue
726 */
727 protected function mergeMagicWords( &$value, $fallbackValue ) {
728 foreach ( $fallbackValue as $magicName => $fallbackInfo ) {
729 if ( !isset( $value[$magicName] ) ) {
730 $value[$magicName] = $fallbackInfo;
731 } else {
732 $oldSynonyms = array_slice( $fallbackInfo, 1 );
733 $newSynonyms = array_slice( $value[$magicName], 1 );
734 $synonyms = array_values( array_unique( array_merge(
735 $newSynonyms, $oldSynonyms ) ) );
736 $value[$magicName] = array_merge( array( $fallbackInfo[0] ), $synonyms );
737 }
738 }
739 }
740
741 /**
742 * Given an array mapping language code to localisation value, such as is
743 * found in extension *.i18n.php files, iterate through a fallback sequence
744 * to merge the given data with an existing primary value.
745 *
746 * Returns true if any data from the extension array was used, false
747 * otherwise.
748 * @param $codeSequence
749 * @param $key
750 * @param $value
751 * @param $fallbackValue
752 * @return bool
753 */
754 protected function mergeExtensionItem( $codeSequence, $key, &$value, $fallbackValue ) {
755 $used = false;
756 foreach ( $codeSequence as $code ) {
757 if ( isset( $fallbackValue[$code] ) ) {
758 $this->mergeItem( $key, $value, $fallbackValue[$code] );
759 $used = true;
760 }
761 }
762
763 return $used;
764 }
765
766 /**
767 * Load localisation data for a given language for both core and extensions
768 * and save it to the persistent cache store and the process cache
769 * @param $code
770 * @throws MWException
771 */
772 public function recache( $code ) {
773 global $wgExtensionMessagesFiles, $wgMessagesDirs;
774 wfProfileIn( __METHOD__ );
775
776 if ( !$code ) {
777 wfProfileOut( __METHOD__ );
778 throw new MWException( "Invalid language code requested" );
779 }
780 $this->recachedLangs[$code] = true;
781
782 # Initial values
783 $initialData = array_combine(
784 self::$allKeys,
785 array_fill( 0, count( self::$allKeys ), null ) );
786 $coreData = $initialData;
787 $deps = array();
788
789 # Load the primary localisation from the source file
790 $data = $this->readSourceFilesAndRegisterDeps( $code, $deps );
791 if ( $data === false ) {
792 wfDebug( __METHOD__ . ": no localisation file for $code, using fallback to en\n" );
793 $coreData['fallback'] = 'en';
794 } else {
795 wfDebug( __METHOD__ . ": got localisation for $code from source\n" );
796
797 # Merge primary localisation
798 foreach ( $data as $key => $value ) {
799 $this->mergeItem( $key, $coreData[$key], $value );
800 }
801 }
802
803 # Fill in the fallback if it's not there already
804 if ( is_null( $coreData['fallback'] ) ) {
805 $coreData['fallback'] = $code === 'en' ? false : 'en';
806 }
807 if ( $coreData['fallback'] === false ) {
808 $coreData['fallbackSequence'] = array();
809 } else {
810 $coreData['fallbackSequence'] = array_map( 'trim', explode( ',', $coreData['fallback'] ) );
811 $len = count( $coreData['fallbackSequence'] );
812
813 # Ensure that the sequence ends at en
814 if ( $coreData['fallbackSequence'][$len - 1] !== 'en' ) {
815 $coreData['fallbackSequence'][] = 'en';
816 }
817
818 # Load the fallback localisation item by item and merge it
819 foreach ( $coreData['fallbackSequence'] as $fbCode ) {
820 # Load the secondary localisation from the source file to
821 # avoid infinite cycles on cyclic fallbacks
822 $fbData = $this->readSourceFilesAndRegisterDeps( $fbCode, $deps );
823 if ( $fbData === false ) {
824 continue;
825 }
826
827 foreach ( self::$allKeys as $key ) {
828 if ( !isset( $fbData[$key] ) ) {
829 continue;
830 }
831
832 if ( is_null( $coreData[$key] ) || $this->isMergeableKey( $key ) ) {
833 $this->mergeItem( $key, $coreData[$key], $fbData[$key] );
834 }
835 }
836 }
837 }
838
839 $codeSequence = array_merge( array( $code ), $coreData['fallbackSequence'] );
840
841 # Load the extension localisations
842 # This is done after the core because we know the fallback sequence now.
843 # But it has a higher precedence for merging so that we can support things
844 # like site-specific message overrides.
845 wfProfileIn( __METHOD__ . '-extensions' );
846 $allData = $initialData;
847 foreach ( $wgMessagesDirs as $dirs ) {
848 foreach ( (array)$dirs as $dir ) {
849 foreach ( $codeSequence as $csCode ) {
850 $fileName = "$dir/$csCode.json";
851 $data = $this->readJSONFile( $fileName );
852
853 foreach ( $data as $key => $item ) {
854 $this->mergeItem( $key, $allData[$key], $item );
855 }
856
857 $deps[] = new FileDependency( $fileName );
858 }
859 }
860 }
861
862 foreach ( $wgExtensionMessagesFiles as $extension => $fileName ) {
863 $data = $this->readPHPFile( $fileName, 'extension' );
864 $used = false;
865
866 foreach ( $data as $key => $item ) {
867 if ( $key === 'messages' && isset( $wgMessagesDirs[$extension] ) ) {
868 # For backwards compatibility, ignore messages from extensions in
869 # $wgExtensionMessagesFiles that are also present in $wgMessagesDirs.
870 # This allows extensions to use both and be backwards compatible.
871 # Variables other than $messages still need to be supported though.
872 continue;
873 }
874 if ( $this->mergeExtensionItem( $codeSequence, $key, $allData[$key], $item ) ) {
875 $used = true;
876 }
877 }
878
879 if ( $used ) {
880 $deps[] = new FileDependency( $fileName );
881 }
882 }
883
884 # Merge core data into extension data
885 foreach ( $coreData as $key => $item ) {
886 $this->mergeItem( $key, $allData[$key], $item );
887 }
888 wfProfileOut( __METHOD__ . '-extensions' );
889
890 # Add cache dependencies for any referenced globals
891 $deps['wgExtensionMessagesFiles'] = new GlobalDependency( 'wgExtensionMessagesFiles' );
892 $deps['wgMessagesDirs'] = new GlobalDependency( 'wgMessagesDirs' );
893 $deps['version'] = new ConstantDependency( 'MW_LC_VERSION' );
894
895 # Add dependencies to the cache entry
896 $allData['deps'] = $deps;
897
898 # Replace spaces with underscores in namespace names
899 $allData['namespaceNames'] = str_replace( ' ', '_', $allData['namespaceNames'] );
900
901 # And do the same for special page aliases. $page is an array.
902 foreach ( $allData['specialPageAliases'] as &$page ) {
903 $page = str_replace( ' ', '_', $page );
904 }
905 # Decouple the reference to prevent accidental damage
906 unset( $page );
907
908 # If there were no plural rules, return an empty array
909 if ( $allData['pluralRules'] === null ) {
910 $allData['pluralRules'] = array();
911 }
912 if ( $allData['compiledPluralRules'] === null ) {
913 $allData['compiledPluralRules'] = array();
914 }
915 # If there were no plural rule types, return an empty array
916 if ( $allData['pluralRuleTypes'] === null ) {
917 $allData['pluralRuleTypes'] = array();
918 }
919
920 # Set the list keys
921 $allData['list'] = array();
922 foreach ( self::$splitKeys as $key ) {
923 $allData['list'][$key] = array_keys( $allData[$key] );
924 }
925 # Run hooks
926 wfRunHooks( 'LocalisationCacheRecache', array( $this, $code, &$allData ) );
927
928 if ( is_null( $allData['namespaceNames'] ) ) {
929 wfProfileOut( __METHOD__ );
930 throw new MWException( __METHOD__ . ': Localisation data failed sanity check! ' .
931 'Check that your languages/messages/MessagesEn.php file is intact.' );
932 }
933
934 # Set the preload key
935 $allData['preload'] = $this->buildPreload( $allData );
936
937 # Save to the process cache and register the items loaded
938 $this->data[$code] = $allData;
939 foreach ( $allData as $key => $item ) {
940 $this->loadedItems[$code][$key] = true;
941 }
942
943 # Save to the persistent cache
944 wfProfileIn( __METHOD__ . '-write' );
945 $this->store->startWrite( $code );
946 foreach ( $allData as $key => $value ) {
947 if ( in_array( $key, self::$splitKeys ) ) {
948 foreach ( $value as $subkey => $subvalue ) {
949 $this->store->set( "$key:$subkey", $subvalue );
950 }
951 } else {
952 $this->store->set( $key, $value );
953 }
954 }
955 $this->store->finishWrite();
956 wfProfileOut( __METHOD__ . '-write' );
957
958 # Clear out the MessageBlobStore
959 # HACK: If using a null (i.e. disabled) storage backend, we
960 # can't write to the MessageBlobStore either
961 if ( !$this->store instanceof LCStoreNull ) {
962 MessageBlobStore::clear();
963 }
964
965 wfProfileOut( __METHOD__ );
966 }
967
968 /**
969 * Build the preload item from the given pre-cache data.
970 *
971 * The preload item will be loaded automatically, improving performance
972 * for the commonly-requested items it contains.
973 * @param $data
974 * @return array
975 */
976 protected function buildPreload( $data ) {
977 $preload = array( 'messages' => array() );
978 foreach ( self::$preloadedKeys as $key ) {
979 $preload[$key] = $data[$key];
980 }
981
982 foreach ( $data['preloadedMessages'] as $subkey ) {
983 if ( isset( $data['messages'][$subkey] ) ) {
984 $subitem = $data['messages'][$subkey];
985 } else {
986 $subitem = null;
987 }
988 $preload['messages'][$subkey] = $subitem;
989 }
990
991 return $preload;
992 }
993
994 /**
995 * Unload the data for a given language from the object cache.
996 * Reduces memory usage.
997 * @param $code
998 */
999 public function unload( $code ) {
1000 unset( $this->data[$code] );
1001 unset( $this->loadedItems[$code] );
1002 unset( $this->loadedSubitems[$code] );
1003 unset( $this->initialisedLangs[$code] );
1004 unset( $this->shallowFallbacks[$code] );
1005
1006 foreach ( $this->shallowFallbacks as $shallowCode => $fbCode ) {
1007 if ( $fbCode === $code ) {
1008 $this->unload( $shallowCode );
1009 }
1010 }
1011 }
1012
1013 /**
1014 * Unload all data
1015 */
1016 public function unloadAll() {
1017 foreach ( $this->initialisedLangs as $lang => $unused ) {
1018 $this->unload( $lang );
1019 }
1020 }
1021
1022 /**
1023 * Disable the storage backend
1024 */
1025 public function disableBackend() {
1026 $this->store = new LCStoreNull;
1027 $this->manualRecache = false;
1028 }
1029 }
1030
1031 /**
1032 * Interface for the persistence layer of LocalisationCache.
1033 *
1034 * The persistence layer is two-level hierarchical cache. The first level
1035 * is the language, the second level is the item or subitem.
1036 *
1037 * Since the data for a whole language is rebuilt in one operation, it needs
1038 * to have a fast and atomic method for deleting or replacing all of the
1039 * current data for a given language. The interface reflects this bulk update
1040 * operation. Callers writing to the cache must first call startWrite(), then
1041 * will call set() a couple of thousand times, then will call finishWrite()
1042 * to commit the operation. When finishWrite() is called, the cache is
1043 * expected to delete all data previously stored for that language.
1044 *
1045 * The values stored are PHP variables suitable for serialize(). Implementations
1046 * of LCStore are responsible for serializing and unserializing.
1047 */
1048 interface LCStore {
1049 /**
1050 * Get a value.
1051 * @param string $code Language code
1052 * @param string $key Cache key
1053 */
1054 function get( $code, $key );
1055
1056 /**
1057 * Start a write transaction.
1058 * @param string $code Language code
1059 */
1060 function startWrite( $code );
1061
1062 /**
1063 * Finish a write transaction.
1064 */
1065 function finishWrite();
1066
1067 /**
1068 * Set a key to a given value. startWrite() must be called before this
1069 * is called, and finishWrite() must be called afterwards.
1070 * @param string $key
1071 * @param mixed $value
1072 */
1073 function set( $key, $value );
1074 }
1075
1076 /**
1077 * LCStore implementation which uses PHP accelerator to store data.
1078 * This will work if one of XCache, WinCache or APC cacher is configured.
1079 * (See ObjectCache.php)
1080 */
1081 class LCStoreAccel implements LCStore {
1082 private $currentLang;
1083 private $keys;
1084
1085 public function __construct() {
1086 $this->cache = wfGetCache( CACHE_ACCEL );
1087 }
1088
1089 public function get( $code, $key ) {
1090 $k = wfMemcKey( 'l10n', $code, 'k', $key );
1091 $r = $this->cache->get( $k );
1092
1093 return $r === false ? null : $r;
1094 }
1095
1096 public function startWrite( $code ) {
1097 $k = wfMemcKey( 'l10n', $code, 'l' );
1098 $keys = $this->cache->get( $k );
1099 if ( $keys ) {
1100 foreach ( $keys as $k ) {
1101 $this->cache->delete( $k );
1102 }
1103 }
1104 $this->currentLang = $code;
1105 $this->keys = array();
1106 }
1107
1108 public function finishWrite() {
1109 if ( $this->currentLang ) {
1110 $k = wfMemcKey( 'l10n', $this->currentLang, 'l' );
1111 $this->cache->set( $k, array_keys( $this->keys ) );
1112 }
1113 $this->currentLang = null;
1114 $this->keys = array();
1115 }
1116
1117 public function set( $key, $value ) {
1118 if ( $this->currentLang ) {
1119 $k = wfMemcKey( 'l10n', $this->currentLang, 'k', $key );
1120 $this->keys[$k] = true;
1121 $this->cache->set( $k, $value );
1122 }
1123 }
1124 }
1125
1126 /**
1127 * LCStore implementation which uses the standard DB functions to store data.
1128 * This will work on any MediaWiki installation.
1129 */
1130 class LCStoreDB implements LCStore {
1131 private $currentLang;
1132 private $writesDone = false;
1133
1134 /**
1135 * @var DatabaseBase
1136 */
1137 private $dbw;
1138 private $batch;
1139 private $readOnly = false;
1140
1141 public function get( $code, $key ) {
1142 if ( $this->writesDone ) {
1143 $db = wfGetDB( DB_MASTER );
1144 } else {
1145 $db = wfGetDB( DB_SLAVE );
1146 }
1147 $row = $db->selectRow( 'l10n_cache', array( 'lc_value' ),
1148 array( 'lc_lang' => $code, 'lc_key' => $key ), __METHOD__ );
1149 if ( $row ) {
1150 return unserialize( $row->lc_value );
1151 } else {
1152 return null;
1153 }
1154 }
1155
1156 public function startWrite( $code ) {
1157 if ( $this->readOnly ) {
1158 return;
1159 }
1160
1161 if ( !$code ) {
1162 throw new MWException( __METHOD__ . ": Invalid language \"$code\"" );
1163 }
1164
1165 $this->dbw = wfGetDB( DB_MASTER );
1166 try {
1167 $this->dbw->begin( __METHOD__ );
1168 $this->dbw->delete( 'l10n_cache', array( 'lc_lang' => $code ), __METHOD__ );
1169 } catch ( DBQueryError $e ) {
1170 if ( $this->dbw->wasReadOnlyError() ) {
1171 $this->readOnly = true;
1172 $this->dbw->rollback( __METHOD__ );
1173
1174 return;
1175 } else {
1176 throw $e;
1177 }
1178 }
1179
1180 $this->currentLang = $code;
1181 $this->batch = array();
1182 }
1183
1184 public function finishWrite() {
1185 if ( $this->readOnly ) {
1186 return;
1187 }
1188
1189 if ( $this->batch ) {
1190 $this->dbw->insert( 'l10n_cache', $this->batch, __METHOD__ );
1191 }
1192
1193 $this->dbw->commit( __METHOD__ );
1194 $this->currentLang = null;
1195 $this->dbw = null;
1196 $this->batch = array();
1197 $this->writesDone = true;
1198 }
1199
1200 public function set( $key, $value ) {
1201 if ( $this->readOnly ) {
1202 return;
1203 }
1204
1205 if ( is_null( $this->currentLang ) ) {
1206 throw new MWException( __CLASS__ . ': must call startWrite() before calling set()' );
1207 }
1208
1209 $this->batch[] = array(
1210 'lc_lang' => $this->currentLang,
1211 'lc_key' => $key,
1212 'lc_value' => serialize( $value ) );
1213
1214 if ( count( $this->batch ) >= 100 ) {
1215 $this->dbw->insert( 'l10n_cache', $this->batch, __METHOD__ );
1216 $this->batch = array();
1217 }
1218 }
1219 }
1220
1221 /**
1222 * LCStore implementation which stores data as a collection of CDB files in the
1223 * directory given by $wgCacheDirectory. If $wgCacheDirectory is not set, this
1224 * will throw an exception.
1225 *
1226 * Profiling indicates that on Linux, this implementation outperforms MySQL if
1227 * the directory is on a local filesystem and there is ample kernel cache
1228 * space. The performance advantage is greater when the DBA extension is
1229 * available than it is with the PHP port.
1230 *
1231 * See Cdb.php and http://cr.yp.to/cdb.html
1232 */
1233 class LCStoreCDB implements LCStore {
1234 private $readers;
1235 private $writer;
1236 private $currentLang;
1237 private $directory;
1238
1239 function __construct( $conf = array() ) {
1240 global $wgCacheDirectory;
1241
1242 if ( isset( $conf['directory'] ) ) {
1243 $this->directory = $conf['directory'];
1244 } else {
1245 $this->directory = $wgCacheDirectory;
1246 }
1247 }
1248
1249 public function get( $code, $key ) {
1250 if ( !isset( $this->readers[$code] ) ) {
1251 $fileName = $this->getFileName( $code );
1252
1253 $this->readers[$code] = false;
1254 if ( file_exists( $fileName ) ) {
1255 try {
1256 $this->readers[$code] = CdbReader::open( $fileName );
1257 } catch ( CdbException $e ) {
1258 wfDebug( __METHOD__ . ": unable to open cdb file for reading" );
1259 }
1260 }
1261 }
1262
1263 if ( !$this->readers[$code] ) {
1264 return null;
1265 } else {
1266 $value = false;
1267 try {
1268 $value = $this->readers[$code]->get( $key );
1269 } catch ( CdbException $e ) {
1270 wfDebug( __METHOD__ . ": CdbException caught, error message was "
1271 . $e->getMessage() );
1272 }
1273 if ( $value === false ) {
1274 return null;
1275 }
1276
1277 return unserialize( $value );
1278 }
1279 }
1280
1281 public function startWrite( $code ) {
1282 if ( !file_exists( $this->directory ) ) {
1283 if ( !wfMkdirParents( $this->directory, null, __METHOD__ ) ) {
1284 throw new MWException( "Unable to create the localisation store " .
1285 "directory \"{$this->directory}\"" );
1286 }
1287 }
1288
1289 // Close reader to stop permission errors on write
1290 if ( !empty( $this->readers[$code] ) ) {
1291 $this->readers[$code]->close();
1292 }
1293
1294 try {
1295 $this->writer = CdbWriter::open( $this->getFileName( $code ) );
1296 } catch ( CdbException $e ) {
1297 throw new MWException( $e->getMessage() );
1298 }
1299 $this->currentLang = $code;
1300 }
1301
1302 public function finishWrite() {
1303 // Close the writer
1304 try {
1305 $this->writer->close();
1306 } catch ( CdbException $e ) {
1307 throw new MWException( $e->getMessage() );
1308 }
1309 $this->writer = null;
1310 unset( $this->readers[$this->currentLang] );
1311 $this->currentLang = null;
1312 }
1313
1314 public function set( $key, $value ) {
1315 if ( is_null( $this->writer ) ) {
1316 throw new MWException( __CLASS__ . ': must call startWrite() before calling set()' );
1317 }
1318 try {
1319 $this->writer->set( $key, serialize( $value ) );
1320 } catch ( CdbException $e ) {
1321 throw new MWException( $e->getMessage() );
1322 }
1323 }
1324
1325 protected function getFileName( $code ) {
1326 if ( strval( $code ) === '' || strpos( $code, '/' ) !== false ) {
1327 throw new MWException( __METHOD__ . ": Invalid language \"$code\"" );
1328 }
1329
1330 return "{$this->directory}/l10n_cache-$code.cdb";
1331 }
1332 }
1333
1334 /**
1335 * Null store backend, used to avoid DB errors during install
1336 */
1337 class LCStoreNull implements LCStore {
1338 public function get( $code, $key ) {
1339 return null;
1340 }
1341
1342 public function startWrite( $code ) {
1343 }
1344
1345 public function finishWrite() {
1346 }
1347
1348 public function set( $key, $value ) {
1349 }
1350 }
1351
1352 /**
1353 * A localisation cache optimised for loading large amounts of data for many
1354 * languages. Used by rebuildLocalisationCache.php.
1355 */
1356 class LocalisationCacheBulkLoad extends LocalisationCache {
1357 /**
1358 * A cache of the contents of data files.
1359 * Core files are serialized to avoid using ~1GB of RAM during a recache.
1360 */
1361 private $fileCache = array();
1362
1363 /**
1364 * Most recently used languages. Uses the linked-list aspect of PHP hashtables
1365 * to keep the most recently used language codes at the end of the array, and
1366 * the language codes that are ready to be deleted at the beginning.
1367 */
1368 private $mruLangs = array();
1369
1370 /**
1371 * Maximum number of languages that may be loaded into $this->data
1372 */
1373 private $maxLoadedLangs = 10;
1374
1375 /**
1376 * @param $fileName
1377 * @param $fileType
1378 * @return array|mixed
1379 */
1380 protected function readPHPFile( $fileName, $fileType ) {
1381 $serialize = $fileType === 'core';
1382 if ( !isset( $this->fileCache[$fileName][$fileType] ) ) {
1383 $data = parent::readPHPFile( $fileName, $fileType );
1384
1385 if ( $serialize ) {
1386 $encData = serialize( $data );
1387 } else {
1388 $encData = $data;
1389 }
1390
1391 $this->fileCache[$fileName][$fileType] = $encData;
1392
1393 return $data;
1394 } elseif ( $serialize ) {
1395 return unserialize( $this->fileCache[$fileName][$fileType] );
1396 } else {
1397 return $this->fileCache[$fileName][$fileType];
1398 }
1399 }
1400
1401 /**
1402 * @param $code
1403 * @param $key
1404 * @return mixed
1405 */
1406 public function getItem( $code, $key ) {
1407 unset( $this->mruLangs[$code] );
1408 $this->mruLangs[$code] = true;
1409
1410 return parent::getItem( $code, $key );
1411 }
1412
1413 /**
1414 * @param $code
1415 * @param $key
1416 * @param $subkey
1417 * @return
1418 */
1419 public function getSubitem( $code, $key, $subkey ) {
1420 unset( $this->mruLangs[$code] );
1421 $this->mruLangs[$code] = true;
1422
1423 return parent::getSubitem( $code, $key, $subkey );
1424 }
1425
1426 /**
1427 * @param $code
1428 */
1429 public function recache( $code ) {
1430 parent::recache( $code );
1431 unset( $this->mruLangs[$code] );
1432 $this->mruLangs[$code] = true;
1433 $this->trimCache();
1434 }
1435
1436 /**
1437 * @param $code
1438 */
1439 public function unload( $code ) {
1440 unset( $this->mruLangs[$code] );
1441 parent::unload( $code );
1442 }
1443
1444 /**
1445 * Unload cached languages until there are less than $this->maxLoadedLangs
1446 */
1447 protected function trimCache() {
1448 while ( count( $this->data ) > $this->maxLoadedLangs && count( $this->mruLangs ) ) {
1449 reset( $this->mruLangs );
1450 $code = key( $this->mruLangs );
1451 wfDebug( __METHOD__ . ": unloading $code\n" );
1452 $this->unload( $code );
1453 }
1454 }
1455 }