Merge "mediawiki.toc.test: Cleanup"
[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 wfDebugLog( 'caches', get_class( $this ) . ": using store $storeClass" );
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 ( 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' );
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 wfSuppressWarnings();
514 $_apcEnabled = ini_set( 'apc.cache_by_default', '0' );
515 wfRestoreWarnings();
516
517 include $_fileName;
518
519 wfSuppressWarnings();
520 ini_set( 'apc.cache_by_default', $_apcEnabled );
521 wfRestoreWarnings();
522
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" );
530 }
531 wfProfileOut( __METHOD__ );
532
533 return $data;
534 }
535
536 /**
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 with a 'messages' key, or empty array if the file doesn't exist
541 */
542 public function readJSONFile( $fileName ) {
543 wfProfileIn( __METHOD__ );
544
545 if ( !is_readable( $fileName ) ) {
546 wfProfileOut( __METHOD__ );
547
548 return array();
549 }
550
551 $json = file_get_contents( $fileName );
552 if ( $json === false ) {
553 wfProfileOut( __METHOD__ );
554
555 return array();
556 }
557
558 $data = FormatJson::decode( $json, true );
559 if ( $data === null ) {
560 wfProfileOut( __METHOD__ );
561
562 throw new MWException( __METHOD__ . ": Invalid JSON file: $fileName" );
563 }
564
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] );
569 }
570 }
571
572 wfProfileOut( __METHOD__ );
573
574 // The JSON format only supports messages, none of the other variables, so wrap the data
575 return array( 'messages' => $data );
576 }
577
578 /**
579 * Get the compiled plural rules for a given language from the XML files.
580 * @since 1.20
581 */
582 public function getCompiledPluralRules( $code ) {
583 $rules = $this->getPluralRules( $code );
584 if ( $rules === null ) {
585 return null;
586 }
587 try {
588 $compiledRules = CLDRPluralRuleEvaluator::compile( $rules );
589 } catch ( CLDRPluralRuleError $e ) {
590 wfDebugLog( 'l10n', $e->getMessage() );
591
592 return array();
593 }
594
595 return $compiledRules;
596 }
597
598 /**
599 * Get the plural rules for a given language from the XML files.
600 * Cached.
601 * @since 1.20
602 */
603 public function getPluralRules( $code ) {
604 if ( $this->pluralRules === null ) {
605 $this->loadPluralFiles();
606 }
607 if ( !isset( $this->pluralRules[$code] ) ) {
608 return null;
609 } else {
610 return $this->pluralRules[$code];
611 }
612 }
613
614 /**
615 * Get the plural rule types for a given language from the XML files.
616 * Cached.
617 * @since 1.22
618 */
619 public function getPluralRuleTypes( $code ) {
620 if ( $this->pluralRuleTypes === null ) {
621 $this->loadPluralFiles();
622 }
623 if ( !isset( $this->pluralRuleTypes[$code] ) ) {
624 return null;
625 } else {
626 return $this->pluralRuleTypes[$code];
627 }
628 }
629
630 /**
631 * Load the plural XML files.
632 */
633 protected function loadPluralFiles() {
634 global $IP;
635 $cldrPlural = "$IP/languages/data/plurals.xml";
636 $mwPlural = "$IP/languages/data/plurals-mediawiki.xml";
637 // Load CLDR plural rules
638 $this->loadPluralFile( $cldrPlural );
639 if ( file_exists( $mwPlural ) ) {
640 // Override or extend
641 $this->loadPluralFile( $mwPlural );
642 }
643 }
644
645 /**
646 * Load a plural XML file with the given filename, compile the relevant
647 * rules, and save the compiled rules in a process-local cache.
648 */
649 protected function loadPluralFile( $fileName ) {
650 $doc = new DOMDocument;
651 $doc->load( $fileName );
652 $rulesets = $doc->getElementsByTagName( "pluralRules" );
653 foreach ( $rulesets as $ruleset ) {
654 $codes = $ruleset->getAttribute( 'locales' );
655 $rules = array();
656 $ruleTypes = array();
657 $ruleElements = $ruleset->getElementsByTagName( "pluralRule" );
658 foreach ( $ruleElements as $elt ) {
659 $ruleType = $elt->getAttribute( 'count' );
660 if ( $ruleType === 'other' ) {
661 // Don't record "other" rules, which have an empty condition
662 continue;
663 }
664 $rules[] = $elt->nodeValue;
665 $ruleTypes[] = $ruleType;
666 }
667 foreach ( explode( ' ', $codes ) as $code ) {
668 $this->pluralRules[$code] = $rules;
669 $this->pluralRuleTypes[$code] = $ruleTypes;
670 }
671 }
672 }
673
674 /**
675 * Read the data from the source files for a given language, and register
676 * the relevant dependencies in the $deps array. If the localisation
677 * exists, the data array is returned, otherwise false is returned.
678 */
679 protected function readSourceFilesAndRegisterDeps( $code, &$deps ) {
680 global $IP;
681 wfProfileIn( __METHOD__ );
682
683
684 // This reads in the PHP i18n file with non-messages l10n data
685 $fileName = Language::getMessagesFileName( $code );
686 if ( !file_exists( $fileName ) ) {
687 $data = array();
688 } else {
689 $deps[] = new FileDependency( $fileName );
690 $data = $this->readPHPFile( $fileName, 'core' );
691 }
692
693 # Load CLDR plural rules for JavaScript
694 $data['pluralRules'] = $this->getPluralRules( $code );
695 # And for PHP
696 $data['compiledPluralRules'] = $this->getCompiledPluralRules( $code );
697 # Load plural rule types
698 $data['pluralRuleTypes'] = $this->getPluralRuleTypes( $code );
699
700 $deps['plurals'] = new FileDependency( "$IP/languages/data/plurals.xml" );
701 $deps['plurals-mw'] = new FileDependency( "$IP/languages/data/plurals-mediawiki.xml" );
702
703 wfProfileOut( __METHOD__ );
704
705 return $data;
706 }
707
708 /**
709 * Merge two localisation values, a primary and a fallback, overwriting the
710 * primary value in place.
711 * @param $key
712 * @param $value
713 * @param $fallbackValue
714 */
715 protected function mergeItem( $key, &$value, $fallbackValue ) {
716 if ( !is_null( $value ) ) {
717 if ( !is_null( $fallbackValue ) ) {
718 if ( in_array( $key, self::$mergeableMapKeys ) ) {
719 $value = $value + $fallbackValue;
720 } elseif ( in_array( $key, self::$mergeableListKeys ) ) {
721 $value = array_unique( array_merge( $fallbackValue, $value ) );
722 } elseif ( in_array( $key, self::$mergeableAliasListKeys ) ) {
723 $value = array_merge_recursive( $value, $fallbackValue );
724 } elseif ( in_array( $key, self::$optionalMergeKeys ) ) {
725 if ( !empty( $value['inherit'] ) ) {
726 $value = array_merge( $fallbackValue, $value );
727 }
728
729 if ( isset( $value['inherit'] ) ) {
730 unset( $value['inherit'] );
731 }
732 } elseif ( in_array( $key, self::$magicWordKeys ) ) {
733 $this->mergeMagicWords( $value, $fallbackValue );
734 }
735 }
736 } else {
737 $value = $fallbackValue;
738 }
739 }
740
741 /**
742 * @param $value
743 * @param $fallbackValue
744 */
745 protected function mergeMagicWords( &$value, $fallbackValue ) {
746 foreach ( $fallbackValue as $magicName => $fallbackInfo ) {
747 if ( !isset( $value[$magicName] ) ) {
748 $value[$magicName] = $fallbackInfo;
749 } else {
750 $oldSynonyms = array_slice( $fallbackInfo, 1 );
751 $newSynonyms = array_slice( $value[$magicName], 1 );
752 $synonyms = array_values( array_unique( array_merge(
753 $newSynonyms, $oldSynonyms ) ) );
754 $value[$magicName] = array_merge( array( $fallbackInfo[0] ), $synonyms );
755 }
756 }
757 }
758
759 /**
760 * Given an array mapping language code to localisation value, such as is
761 * found in extension *.i18n.php files, iterate through a fallback sequence
762 * to merge the given data with an existing primary value.
763 *
764 * Returns true if any data from the extension array was used, false
765 * otherwise.
766 * @param $codeSequence
767 * @param $key
768 * @param $value
769 * @param $fallbackValue
770 * @return bool
771 */
772 protected function mergeExtensionItem( $codeSequence, $key, &$value, $fallbackValue ) {
773 $used = false;
774 foreach ( $codeSequence as $code ) {
775 if ( isset( $fallbackValue[$code] ) ) {
776 $this->mergeItem( $key, $value, $fallbackValue[$code] );
777 $used = true;
778 }
779 }
780
781 return $used;
782 }
783
784 /**
785 * Load localisation data for a given language for both core and extensions
786 * and save it to the persistent cache store and the process cache
787 * @param $code
788 * @throws MWException
789 */
790 public function recache( $code ) {
791 global $wgExtensionMessagesFiles, $wgMessagesDirs;
792 wfProfileIn( __METHOD__ );
793
794 if ( !$code ) {
795 wfProfileOut( __METHOD__ );
796 throw new MWException( "Invalid language code requested" );
797 }
798 $this->recachedLangs[$code] = true;
799
800 # Initial values
801 $initialData = array_combine(
802 self::$allKeys,
803 array_fill( 0, count( self::$allKeys ), null ) );
804 $coreData = $initialData;
805 $deps = array();
806
807 # Load the primary localisation from the source file
808 $data = $this->readSourceFilesAndRegisterDeps( $code, $deps );
809 if ( $data === false ) {
810 wfDebug( __METHOD__ . ": no localisation file for $code, using fallback to en\n" );
811 $coreData['fallback'] = 'en';
812 } else {
813 wfDebug( __METHOD__ . ": got localisation for $code from source\n" );
814
815 # Merge primary localisation
816 foreach ( $data as $key => $value ) {
817 $this->mergeItem( $key, $coreData[$key], $value );
818 }
819 }
820
821 # Fill in the fallback if it's not there already
822 if ( is_null( $coreData['fallback'] ) ) {
823 $coreData['fallback'] = $code === 'en' ? false : 'en';
824 }
825 if ( $coreData['fallback'] === false ) {
826 $coreData['fallbackSequence'] = array();
827 } else {
828 $coreData['fallbackSequence'] = array_map( 'trim', explode( ',', $coreData['fallback'] ) );
829 $len = count( $coreData['fallbackSequence'] );
830
831 # Ensure that the sequence ends at en
832 if ( $coreData['fallbackSequence'][$len - 1] !== 'en' ) {
833 $coreData['fallbackSequence'][] = 'en';
834 }
835
836 # Load the fallback localisation item by item and merge it
837 foreach ( $coreData['fallbackSequence'] as $fbCode ) {
838 # Load the secondary localisation from the source file to
839 # avoid infinite cycles on cyclic fallbacks
840 $fbData = $this->readSourceFilesAndRegisterDeps( $fbCode, $deps );
841 if ( $fbData === false ) {
842 continue;
843 }
844
845 foreach ( self::$allKeys as $key ) {
846 if ( !isset( $fbData[$key] ) ) {
847 continue;
848 }
849
850 if ( is_null( $coreData[$key] ) || $this->isMergeableKey( $key ) ) {
851 $this->mergeItem( $key, $coreData[$key], $fbData[$key] );
852 }
853 }
854 }
855 }
856
857 $codeSequence = array_merge( array( $code ), $coreData['fallbackSequence'] );
858
859 # Load core messages and the extension localisations.
860 wfProfileIn( __METHOD__ . '-extensions' );
861 $allData = $initialData;
862 foreach ( $wgMessagesDirs as $dirs ) {
863 foreach ( (array)$dirs as $dir ) {
864 foreach ( $codeSequence as $csCode ) {
865 $fileName = "$dir/$csCode.json";
866 $data = $this->readJSONFile( $fileName );
867
868 foreach ( $data as $key => $item ) {
869 $this->mergeItem( $key, $allData[$key], $item );
870 }
871
872 $deps[] = new FileDependency( $fileName );
873 }
874 }
875 }
876
877 foreach ( $wgExtensionMessagesFiles as $extension => $fileName ) {
878 if ( isset( $wgMessagesDirs[$extension] ) ) {
879 # Already loaded the JSON files for this extension; skip the PHP shim
880 continue;
881 }
882
883 $data = $this->readPHPFile( $fileName, 'extension' );
884 $used = false;
885
886 foreach ( $data as $key => $item ) {
887 if ( $this->mergeExtensionItem( $codeSequence, $key, $allData[$key], $item ) ) {
888 $used = true;
889 }
890 }
891
892 if ( $used ) {
893 $deps[] = new FileDependency( $fileName );
894 }
895 }
896
897 # Merge core data into extension data
898 foreach ( $coreData as $key => $item ) {
899 $this->mergeItem( $key, $allData[$key], $item );
900 }
901 wfProfileOut( __METHOD__ . '-extensions' );
902
903 # Add cache dependencies for any referenced globals
904 $deps['wgExtensionMessagesFiles'] = new GlobalDependency( 'wgExtensionMessagesFiles' );
905 $deps['wgMessagesDirs'] = new GlobalDependency( 'wgMessagesDirs' );
906 $deps['version'] = new ConstantDependency( 'MW_LC_VERSION' );
907
908 # Add dependencies to the cache entry
909 $allData['deps'] = $deps;
910
911 # Replace spaces with underscores in namespace names
912 $allData['namespaceNames'] = str_replace( ' ', '_', $allData['namespaceNames'] );
913
914 # And do the same for special page aliases. $page is an array.
915 foreach ( $allData['specialPageAliases'] as &$page ) {
916 $page = str_replace( ' ', '_', $page );
917 }
918 # Decouple the reference to prevent accidental damage
919 unset( $page );
920
921 # If there were no plural rules, return an empty array
922 if ( $allData['pluralRules'] === null ) {
923 $allData['pluralRules'] = array();
924 }
925 if ( $allData['compiledPluralRules'] === null ) {
926 $allData['compiledPluralRules'] = array();
927 }
928 # If there were no plural rule types, return an empty array
929 if ( $allData['pluralRuleTypes'] === null ) {
930 $allData['pluralRuleTypes'] = array();
931 }
932
933 # Set the list keys
934 $allData['list'] = array();
935 foreach ( self::$splitKeys as $key ) {
936 $allData['list'][$key] = array_keys( $allData[$key] );
937 }
938 # Run hooks
939 $purgeBlobs = true;
940 wfRunHooks( 'LocalisationCacheRecache', array( $this, $code, &$allData, &$purgeBlobs ) );
941
942 if ( is_null( $allData['namespaceNames'] ) ) {
943 wfProfileOut( __METHOD__ );
944 throw new MWException( __METHOD__ . ': Localisation data failed sanity check! ' .
945 'Check that your languages/messages/MessagesEn.php file is intact.' );
946 }
947
948 # Set the preload key
949 $allData['preload'] = $this->buildPreload( $allData );
950
951 # Save to the process cache and register the items loaded
952 $this->data[$code] = $allData;
953 foreach ( $allData as $key => $item ) {
954 $this->loadedItems[$code][$key] = true;
955 }
956
957 # Save to the persistent cache
958 wfProfileIn( __METHOD__ . '-write' );
959 $this->store->startWrite( $code );
960 foreach ( $allData as $key => $value ) {
961 if ( in_array( $key, self::$splitKeys ) ) {
962 foreach ( $value as $subkey => $subvalue ) {
963 $this->store->set( "$key:$subkey", $subvalue );
964 }
965 } else {
966 $this->store->set( $key, $value );
967 }
968 }
969 $this->store->finishWrite();
970 wfProfileOut( __METHOD__ . '-write' );
971
972 # Clear out the MessageBlobStore
973 # HACK: If using a null (i.e. disabled) storage backend, we
974 # can't write to the MessageBlobStore either
975 if ( $purgeBlobs && !$this->store instanceof LCStoreNull ) {
976 MessageBlobStore::clear();
977 }
978
979 wfProfileOut( __METHOD__ );
980 }
981
982 /**
983 * Build the preload item from the given pre-cache data.
984 *
985 * The preload item will be loaded automatically, improving performance
986 * for the commonly-requested items it contains.
987 * @param $data
988 * @return array
989 */
990 protected function buildPreload( $data ) {
991 $preload = array( 'messages' => array() );
992 foreach ( self::$preloadedKeys as $key ) {
993 $preload[$key] = $data[$key];
994 }
995
996 foreach ( $data['preloadedMessages'] as $subkey ) {
997 if ( isset( $data['messages'][$subkey] ) ) {
998 $subitem = $data['messages'][$subkey];
999 } else {
1000 $subitem = null;
1001 }
1002 $preload['messages'][$subkey] = $subitem;
1003 }
1004
1005 return $preload;
1006 }
1007
1008 /**
1009 * Unload the data for a given language from the object cache.
1010 * Reduces memory usage.
1011 * @param $code
1012 */
1013 public function unload( $code ) {
1014 unset( $this->data[$code] );
1015 unset( $this->loadedItems[$code] );
1016 unset( $this->loadedSubitems[$code] );
1017 unset( $this->initialisedLangs[$code] );
1018 unset( $this->shallowFallbacks[$code] );
1019
1020 foreach ( $this->shallowFallbacks as $shallowCode => $fbCode ) {
1021 if ( $fbCode === $code ) {
1022 $this->unload( $shallowCode );
1023 }
1024 }
1025 }
1026
1027 /**
1028 * Unload all data
1029 */
1030 public function unloadAll() {
1031 foreach ( $this->initialisedLangs as $lang => $unused ) {
1032 $this->unload( $lang );
1033 }
1034 }
1035
1036 /**
1037 * Disable the storage backend
1038 */
1039 public function disableBackend() {
1040 $this->store = new LCStoreNull;
1041 $this->manualRecache = false;
1042 }
1043 }
1044
1045 /**
1046 * Interface for the persistence layer of LocalisationCache.
1047 *
1048 * The persistence layer is two-level hierarchical cache. The first level
1049 * is the language, the second level is the item or subitem.
1050 *
1051 * Since the data for a whole language is rebuilt in one operation, it needs
1052 * to have a fast and atomic method for deleting or replacing all of the
1053 * current data for a given language. The interface reflects this bulk update
1054 * operation. Callers writing to the cache must first call startWrite(), then
1055 * will call set() a couple of thousand times, then will call finishWrite()
1056 * to commit the operation. When finishWrite() is called, the cache is
1057 * expected to delete all data previously stored for that language.
1058 *
1059 * The values stored are PHP variables suitable for serialize(). Implementations
1060 * of LCStore are responsible for serializing and unserializing.
1061 */
1062 interface LCStore {
1063 /**
1064 * Get a value.
1065 * @param string $code Language code
1066 * @param string $key Cache key
1067 */
1068 function get( $code, $key );
1069
1070 /**
1071 * Start a write transaction.
1072 * @param string $code Language code
1073 */
1074 function startWrite( $code );
1075
1076 /**
1077 * Finish a write transaction.
1078 */
1079 function finishWrite();
1080
1081 /**
1082 * Set a key to a given value. startWrite() must be called before this
1083 * is called, and finishWrite() must be called afterwards.
1084 * @param string $key
1085 * @param mixed $value
1086 */
1087 function set( $key, $value );
1088 }
1089
1090 /**
1091 * LCStore implementation which uses PHP accelerator to store data.
1092 * This will work if one of XCache, WinCache or APC cacher is configured.
1093 * (See ObjectCache.php)
1094 */
1095 class LCStoreAccel implements LCStore {
1096 private $currentLang;
1097 private $keys;
1098
1099 public function __construct() {
1100 $this->cache = wfGetCache( CACHE_ACCEL );
1101 }
1102
1103 public function get( $code, $key ) {
1104 $k = wfMemcKey( 'l10n', $code, 'k', $key );
1105 $r = $this->cache->get( $k );
1106
1107 return $r === false ? null : $r;
1108 }
1109
1110 public function startWrite( $code ) {
1111 $k = wfMemcKey( 'l10n', $code, 'l' );
1112 $keys = $this->cache->get( $k );
1113 if ( $keys ) {
1114 foreach ( $keys as $k ) {
1115 $this->cache->delete( $k );
1116 }
1117 }
1118 $this->currentLang = $code;
1119 $this->keys = array();
1120 }
1121
1122 public function finishWrite() {
1123 if ( $this->currentLang ) {
1124 $k = wfMemcKey( 'l10n', $this->currentLang, 'l' );
1125 $this->cache->set( $k, array_keys( $this->keys ) );
1126 }
1127 $this->currentLang = null;
1128 $this->keys = array();
1129 }
1130
1131 public function set( $key, $value ) {
1132 if ( $this->currentLang ) {
1133 $k = wfMemcKey( 'l10n', $this->currentLang, 'k', $key );
1134 $this->keys[$k] = true;
1135 $this->cache->set( $k, $value );
1136 }
1137 }
1138 }
1139
1140 /**
1141 * LCStore implementation which uses the standard DB functions to store data.
1142 * This will work on any MediaWiki installation.
1143 */
1144 class LCStoreDB implements LCStore {
1145 private $currentLang;
1146 private $writesDone = false;
1147
1148 /**
1149 * @var DatabaseBase
1150 */
1151 private $dbw;
1152 private $batch;
1153 private $readOnly = false;
1154
1155 public function get( $code, $key ) {
1156 if ( $this->writesDone ) {
1157 $db = wfGetDB( DB_MASTER );
1158 } else {
1159 $db = wfGetDB( DB_SLAVE );
1160 }
1161 $row = $db->selectRow( 'l10n_cache', array( 'lc_value' ),
1162 array( 'lc_lang' => $code, 'lc_key' => $key ), __METHOD__ );
1163 if ( $row ) {
1164 return unserialize( $row->lc_value );
1165 } else {
1166 return null;
1167 }
1168 }
1169
1170 public function startWrite( $code ) {
1171 if ( $this->readOnly ) {
1172 return;
1173 }
1174
1175 if ( !$code ) {
1176 throw new MWException( __METHOD__ . ": Invalid language \"$code\"" );
1177 }
1178
1179 $this->dbw = wfGetDB( DB_MASTER );
1180 try {
1181 $this->dbw->begin( __METHOD__ );
1182 $this->dbw->delete( 'l10n_cache', array( 'lc_lang' => $code ), __METHOD__ );
1183 } catch ( DBQueryError $e ) {
1184 if ( $this->dbw->wasReadOnlyError() ) {
1185 $this->readOnly = true;
1186 $this->dbw->rollback( __METHOD__ );
1187
1188 return;
1189 } else {
1190 throw $e;
1191 }
1192 }
1193
1194 $this->currentLang = $code;
1195 $this->batch = array();
1196 }
1197
1198 public function finishWrite() {
1199 if ( $this->readOnly ) {
1200 return;
1201 }
1202
1203 if ( $this->batch ) {
1204 $this->dbw->insert( 'l10n_cache', $this->batch, __METHOD__ );
1205 }
1206
1207 $this->dbw->commit( __METHOD__ );
1208 $this->currentLang = null;
1209 $this->dbw = null;
1210 $this->batch = array();
1211 $this->writesDone = true;
1212 }
1213
1214 public function set( $key, $value ) {
1215 if ( $this->readOnly ) {
1216 return;
1217 }
1218
1219 if ( is_null( $this->currentLang ) ) {
1220 throw new MWException( __CLASS__ . ': must call startWrite() before calling set()' );
1221 }
1222
1223 $this->batch[] = array(
1224 'lc_lang' => $this->currentLang,
1225 'lc_key' => $key,
1226 'lc_value' => serialize( $value ) );
1227
1228 if ( count( $this->batch ) >= 100 ) {
1229 $this->dbw->insert( 'l10n_cache', $this->batch, __METHOD__ );
1230 $this->batch = array();
1231 }
1232 }
1233 }
1234
1235 /**
1236 * LCStore implementation which stores data as a collection of CDB files in the
1237 * directory given by $wgCacheDirectory. If $wgCacheDirectory is not set, this
1238 * will throw an exception.
1239 *
1240 * Profiling indicates that on Linux, this implementation outperforms MySQL if
1241 * the directory is on a local filesystem and there is ample kernel cache
1242 * space. The performance advantage is greater when the DBA extension is
1243 * available than it is with the PHP port.
1244 *
1245 * See Cdb.php and http://cr.yp.to/cdb.html
1246 */
1247 class LCStoreCDB implements LCStore {
1248 /** @var CdbReader[] */
1249 private $readers;
1250
1251 /** @var CdbWriter */
1252 private $writer;
1253
1254 /** @var string Current language code */
1255 private $currentLang;
1256
1257 /** @var bool|string Cache directory. False if not set */
1258 private $directory;
1259
1260 function __construct( $conf = array() ) {
1261 global $wgCacheDirectory;
1262
1263 if ( isset( $conf['directory'] ) ) {
1264 $this->directory = $conf['directory'];
1265 } else {
1266 $this->directory = $wgCacheDirectory;
1267 }
1268 }
1269
1270 public function get( $code, $key ) {
1271 if ( !isset( $this->readers[$code] ) ) {
1272 $fileName = $this->getFileName( $code );
1273
1274 $this->readers[$code] = false;
1275 if ( file_exists( $fileName ) ) {
1276 try {
1277 $this->readers[$code] = CdbReader::open( $fileName );
1278 } catch ( CdbException $e ) {
1279 wfDebug( __METHOD__ . ": unable to open cdb file for reading\n" );
1280 }
1281 }
1282 }
1283
1284 if ( !$this->readers[$code] ) {
1285 return null;
1286 } else {
1287 $value = false;
1288 try {
1289 $value = $this->readers[$code]->get( $key );
1290 } catch ( CdbException $e ) {
1291 wfDebug( __METHOD__ . ": CdbException caught, error message was "
1292 . $e->getMessage() . "\n" );
1293 }
1294 if ( $value === false ) {
1295 return null;
1296 }
1297
1298 return unserialize( $value );
1299 }
1300 }
1301
1302 public function startWrite( $code ) {
1303 if ( !file_exists( $this->directory ) ) {
1304 if ( !wfMkdirParents( $this->directory, null, __METHOD__ ) ) {
1305 throw new MWException( "Unable to create the localisation store " .
1306 "directory \"{$this->directory}\"" );
1307 }
1308 }
1309
1310 // Close reader to stop permission errors on write
1311 if ( !empty( $this->readers[$code] ) ) {
1312 $this->readers[$code]->close();
1313 }
1314
1315 try {
1316 $this->writer = CdbWriter::open( $this->getFileName( $code ) );
1317 } catch ( CdbException $e ) {
1318 throw new MWException( $e->getMessage() );
1319 }
1320 $this->currentLang = $code;
1321 }
1322
1323 public function finishWrite() {
1324 // Close the writer
1325 try {
1326 $this->writer->close();
1327 } catch ( CdbException $e ) {
1328 throw new MWException( $e->getMessage() );
1329 }
1330 $this->writer = null;
1331 unset( $this->readers[$this->currentLang] );
1332 $this->currentLang = null;
1333 }
1334
1335 public function set( $key, $value ) {
1336 if ( is_null( $this->writer ) ) {
1337 throw new MWException( __CLASS__ . ': must call startWrite() before calling set()' );
1338 }
1339 try {
1340 $this->writer->set( $key, serialize( $value ) );
1341 } catch ( CdbException $e ) {
1342 throw new MWException( $e->getMessage() );
1343 }
1344 }
1345
1346 protected function getFileName( $code ) {
1347 if ( strval( $code ) === '' || strpos( $code, '/' ) !== false ) {
1348 throw new MWException( __METHOD__ . ": Invalid language \"$code\"" );
1349 }
1350
1351 return "{$this->directory}/l10n_cache-$code.cdb";
1352 }
1353 }
1354
1355 /**
1356 * Null store backend, used to avoid DB errors during install
1357 */
1358 class LCStoreNull implements LCStore {
1359 public function get( $code, $key ) {
1360 return null;
1361 }
1362
1363 public function startWrite( $code ) {
1364 }
1365
1366 public function finishWrite() {
1367 }
1368
1369 public function set( $key, $value ) {
1370 }
1371 }
1372
1373 /**
1374 * A localisation cache optimised for loading large amounts of data for many
1375 * languages. Used by rebuildLocalisationCache.php.
1376 */
1377 class LocalisationCacheBulkLoad extends LocalisationCache {
1378 /**
1379 * A cache of the contents of data files.
1380 * Core files are serialized to avoid using ~1GB of RAM during a recache.
1381 */
1382 private $fileCache = array();
1383
1384 /**
1385 * Most recently used languages. Uses the linked-list aspect of PHP hashtables
1386 * to keep the most recently used language codes at the end of the array, and
1387 * the language codes that are ready to be deleted at the beginning.
1388 */
1389 private $mruLangs = array();
1390
1391 /**
1392 * Maximum number of languages that may be loaded into $this->data
1393 */
1394 private $maxLoadedLangs = 10;
1395
1396 /**
1397 * @param $fileName
1398 * @param $fileType
1399 * @return array|mixed
1400 */
1401 protected function readPHPFile( $fileName, $fileType ) {
1402 $serialize = $fileType === 'core';
1403 if ( !isset( $this->fileCache[$fileName][$fileType] ) ) {
1404 $data = parent::readPHPFile( $fileName, $fileType );
1405
1406 if ( $serialize ) {
1407 $encData = serialize( $data );
1408 } else {
1409 $encData = $data;
1410 }
1411
1412 $this->fileCache[$fileName][$fileType] = $encData;
1413
1414 return $data;
1415 } elseif ( $serialize ) {
1416 return unserialize( $this->fileCache[$fileName][$fileType] );
1417 } else {
1418 return $this->fileCache[$fileName][$fileType];
1419 }
1420 }
1421
1422 /**
1423 * @param $code
1424 * @param $key
1425 * @return mixed
1426 */
1427 public function getItem( $code, $key ) {
1428 unset( $this->mruLangs[$code] );
1429 $this->mruLangs[$code] = true;
1430
1431 return parent::getItem( $code, $key );
1432 }
1433
1434 /**
1435 * @param $code
1436 * @param $key
1437 * @param $subkey
1438 * @return
1439 */
1440 public function getSubitem( $code, $key, $subkey ) {
1441 unset( $this->mruLangs[$code] );
1442 $this->mruLangs[$code] = true;
1443
1444 return parent::getSubitem( $code, $key, $subkey );
1445 }
1446
1447 /**
1448 * @param $code
1449 */
1450 public function recache( $code ) {
1451 parent::recache( $code );
1452 unset( $this->mruLangs[$code] );
1453 $this->mruLangs[$code] = true;
1454 $this->trimCache();
1455 }
1456
1457 /**
1458 * @param $code
1459 */
1460 public function unload( $code ) {
1461 unset( $this->mruLangs[$code] );
1462 parent::unload( $code );
1463 }
1464
1465 /**
1466 * Unload cached languages until there are less than $this->maxLoadedLangs
1467 */
1468 protected function trimCache() {
1469 while ( count( $this->data ) > $this->maxLoadedLangs && count( $this->mruLangs ) ) {
1470 reset( $this->mruLangs );
1471 $code = key( $this->mruLangs );
1472 wfDebug( __METHOD__ . ": unloading $code\n" );
1473 $this->unload( $code );
1474 }
1475 }
1476 }