Merge "Add PreferencesFormPreSave hook"
[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 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 protected function readJSONFile( $fileName ) {
543 wfProfileIn( __METHOD__ );
544 if ( !is_readable( $fileName ) ) {
545 return array();
546 }
547
548 $json = file_get_contents( $fileName );
549 if ( $json === false ) {
550 return array();
551 }
552 $data = FormatJson::decode( $json, true );
553 if ( $data === null ) {
554 throw new MWException( __METHOD__ . ": Invalid JSON file: $fileName" );
555 }
556 // Remove keys starting with '@', they're reserved for metadata and non-message data
557 foreach ( $data as $key => $unused ) {
558 if ( $key === '' || $key[0] === '@' ) {
559 unset( $data[$key] );
560 }
561 }
562 // The JSON format only supports messages, none of the other variables, so wrap the data
563 return array( 'messages' => $data );
564 }
565
566 /**
567 * Get the compiled plural rules for a given language from the XML files.
568 * @since 1.20
569 */
570 public function getCompiledPluralRules( $code ) {
571 $rules = $this->getPluralRules( $code );
572 if ( $rules === null ) {
573 return null;
574 }
575 try {
576 $compiledRules = CLDRPluralRuleEvaluator::compile( $rules );
577 } catch ( CLDRPluralRuleError $e ) {
578 wfDebugLog( 'l10n', $e->getMessage() . "\n" );
579
580 return array();
581 }
582
583 return $compiledRules;
584 }
585
586 /**
587 * Get the plural rules for a given language from the XML files.
588 * Cached.
589 * @since 1.20
590 */
591 public function getPluralRules( $code ) {
592 if ( $this->pluralRules === null ) {
593 $this->loadPluralFiles();
594 }
595 if ( !isset( $this->pluralRules[$code] ) ) {
596 return null;
597 } else {
598 return $this->pluralRules[$code];
599 }
600 }
601
602 /**
603 * Get the plural rule types for a given language from the XML files.
604 * Cached.
605 * @since 1.22
606 */
607 public function getPluralRuleTypes( $code ) {
608 if ( $this->pluralRuleTypes === null ) {
609 $this->loadPluralFiles();
610 }
611 if ( !isset( $this->pluralRuleTypes[$code] ) ) {
612 return null;
613 } else {
614 return $this->pluralRuleTypes[$code];
615 }
616 }
617
618 /**
619 * Load the plural XML files.
620 */
621 protected function loadPluralFiles() {
622 global $IP;
623 $cldrPlural = "$IP/languages/data/plurals.xml";
624 $mwPlural = "$IP/languages/data/plurals-mediawiki.xml";
625 // Load CLDR plural rules
626 $this->loadPluralFile( $cldrPlural );
627 if ( file_exists( $mwPlural ) ) {
628 // Override or extend
629 $this->loadPluralFile( $mwPlural );
630 }
631 }
632
633 /**
634 * Load a plural XML file with the given filename, compile the relevant
635 * rules, and save the compiled rules in a process-local cache.
636 */
637 protected function loadPluralFile( $fileName ) {
638 $doc = new DOMDocument;
639 $doc->load( $fileName );
640 $rulesets = $doc->getElementsByTagName( "pluralRules" );
641 foreach ( $rulesets as $ruleset ) {
642 $codes = $ruleset->getAttribute( 'locales' );
643 $rules = array();
644 $ruleTypes = array();
645 $ruleElements = $ruleset->getElementsByTagName( "pluralRule" );
646 foreach ( $ruleElements as $elt ) {
647 $ruleType = $elt->getAttribute( 'count' );
648 if ( $ruleType === 'other' ) {
649 // Don't record "other" rules, which have an empty condition
650 continue;
651 }
652 $rules[] = $elt->nodeValue;
653 $ruleTypes[] = $ruleType;
654 }
655 foreach ( explode( ' ', $codes ) as $code ) {
656 $this->pluralRules[$code] = $rules;
657 $this->pluralRuleTypes[$code] = $ruleTypes;
658 }
659 }
660 }
661
662 /**
663 * Read the data from the source files for a given language, and register
664 * the relevant dependencies in the $deps array. If the localisation
665 * exists, the data array is returned, otherwise false is returned.
666 */
667 protected function readSourceFilesAndRegisterDeps( $code, &$deps ) {
668 global $IP;
669 wfProfileIn( __METHOD__ );
670
671 $fileName = Language::getMessagesFileName( $code );
672 if ( !file_exists( $fileName ) ) {
673 wfProfileOut( __METHOD__ );
674
675 return false;
676 }
677
678 $deps[] = new FileDependency( $fileName );
679 $data = $this->readPHPFile( $fileName, 'core' );
680
681 # Load CLDR plural rules for JavaScript
682 $data['pluralRules'] = $this->getPluralRules( $code );
683 # And for PHP
684 $data['compiledPluralRules'] = $this->getCompiledPluralRules( $code );
685 # Load plural rule types
686 $data['pluralRuleTypes'] = $this->getPluralRuleTypes( $code );
687
688 $deps['plurals'] = new FileDependency( "$IP/languages/data/plurals.xml" );
689 $deps['plurals-mw'] = new FileDependency( "$IP/languages/data/plurals-mediawiki.xml" );
690
691 wfProfileOut( __METHOD__ );
692
693 return $data;
694 }
695
696 /**
697 * Merge two localisation values, a primary and a fallback, overwriting the
698 * primary value in place.
699 * @param $key
700 * @param $value
701 * @param $fallbackValue
702 */
703 protected function mergeItem( $key, &$value, $fallbackValue ) {
704 if ( !is_null( $value ) ) {
705 if ( !is_null( $fallbackValue ) ) {
706 if ( in_array( $key, self::$mergeableMapKeys ) ) {
707 $value = $value + $fallbackValue;
708 } elseif ( in_array( $key, self::$mergeableListKeys ) ) {
709 $value = array_unique( array_merge( $fallbackValue, $value ) );
710 } elseif ( in_array( $key, self::$mergeableAliasListKeys ) ) {
711 $value = array_merge_recursive( $value, $fallbackValue );
712 } elseif ( in_array( $key, self::$optionalMergeKeys ) ) {
713 if ( !empty( $value['inherit'] ) ) {
714 $value = array_merge( $fallbackValue, $value );
715 }
716
717 if ( isset( $value['inherit'] ) ) {
718 unset( $value['inherit'] );
719 }
720 } elseif ( in_array( $key, self::$magicWordKeys ) ) {
721 $this->mergeMagicWords( $value, $fallbackValue );
722 }
723 }
724 } else {
725 $value = $fallbackValue;
726 }
727 }
728
729 /**
730 * @param $value
731 * @param $fallbackValue
732 */
733 protected function mergeMagicWords( &$value, $fallbackValue ) {
734 foreach ( $fallbackValue as $magicName => $fallbackInfo ) {
735 if ( !isset( $value[$magicName] ) ) {
736 $value[$magicName] = $fallbackInfo;
737 } else {
738 $oldSynonyms = array_slice( $fallbackInfo, 1 );
739 $newSynonyms = array_slice( $value[$magicName], 1 );
740 $synonyms = array_values( array_unique( array_merge(
741 $newSynonyms, $oldSynonyms ) ) );
742 $value[$magicName] = array_merge( array( $fallbackInfo[0] ), $synonyms );
743 }
744 }
745 }
746
747 /**
748 * Given an array mapping language code to localisation value, such as is
749 * found in extension *.i18n.php files, iterate through a fallback sequence
750 * to merge the given data with an existing primary value.
751 *
752 * Returns true if any data from the extension array was used, false
753 * otherwise.
754 * @param $codeSequence
755 * @param $key
756 * @param $value
757 * @param $fallbackValue
758 * @return bool
759 */
760 protected function mergeExtensionItem( $codeSequence, $key, &$value, $fallbackValue ) {
761 $used = false;
762 foreach ( $codeSequence as $code ) {
763 if ( isset( $fallbackValue[$code] ) ) {
764 $this->mergeItem( $key, $value, $fallbackValue[$code] );
765 $used = true;
766 }
767 }
768
769 return $used;
770 }
771
772 /**
773 * Load localisation data for a given language for both core and extensions
774 * and save it to the persistent cache store and the process cache
775 * @param $code
776 * @throws MWException
777 */
778 public function recache( $code ) {
779 global $wgExtensionMessagesFiles, $wgMessagesDirs;
780 wfProfileIn( __METHOD__ );
781
782 if ( !$code ) {
783 wfProfileOut( __METHOD__ );
784 throw new MWException( "Invalid language code requested" );
785 }
786 $this->recachedLangs[$code] = true;
787
788 # Initial values
789 $initialData = array_combine(
790 self::$allKeys,
791 array_fill( 0, count( self::$allKeys ), null ) );
792 $coreData = $initialData;
793 $deps = array();
794
795 # Load the primary localisation from the source file
796 $data = $this->readSourceFilesAndRegisterDeps( $code, $deps );
797 if ( $data === false ) {
798 wfDebug( __METHOD__ . ": no localisation file for $code, using fallback to en\n" );
799 $coreData['fallback'] = 'en';
800 } else {
801 wfDebug( __METHOD__ . ": got localisation for $code from source\n" );
802
803 # Merge primary localisation
804 foreach ( $data as $key => $value ) {
805 $this->mergeItem( $key, $coreData[$key], $value );
806 }
807 }
808
809 # Fill in the fallback if it's not there already
810 if ( is_null( $coreData['fallback'] ) ) {
811 $coreData['fallback'] = $code === 'en' ? false : 'en';
812 }
813 if ( $coreData['fallback'] === false ) {
814 $coreData['fallbackSequence'] = array();
815 } else {
816 $coreData['fallbackSequence'] = array_map( 'trim', explode( ',', $coreData['fallback'] ) );
817 $len = count( $coreData['fallbackSequence'] );
818
819 # Ensure that the sequence ends at en
820 if ( $coreData['fallbackSequence'][$len - 1] !== 'en' ) {
821 $coreData['fallbackSequence'][] = 'en';
822 }
823
824 # Load the fallback localisation item by item and merge it
825 foreach ( $coreData['fallbackSequence'] as $fbCode ) {
826 # Load the secondary localisation from the source file to
827 # avoid infinite cycles on cyclic fallbacks
828 $fbData = $this->readSourceFilesAndRegisterDeps( $fbCode, $deps );
829 if ( $fbData === false ) {
830 continue;
831 }
832
833 foreach ( self::$allKeys as $key ) {
834 if ( !isset( $fbData[$key] ) ) {
835 continue;
836 }
837
838 if ( is_null( $coreData[$key] ) || $this->isMergeableKey( $key ) ) {
839 $this->mergeItem( $key, $coreData[$key], $fbData[$key] );
840 }
841 }
842 }
843 }
844
845 $codeSequence = array_merge( array( $code ), $coreData['fallbackSequence'] );
846
847 # Load the extension localisations
848 # This is done after the core because we know the fallback sequence now.
849 # But it has a higher precedence for merging so that we can support things
850 # like site-specific message overrides.
851 wfProfileIn( __METHOD__ . '-extensions' );
852 $allData = $initialData;
853 foreach ( $wgMessagesDirs as $dirs ) {
854 foreach ( (array)$dirs as $dir ) {
855 foreach ( $codeSequence as $csCode ) {
856 $fileName = "$dir/$csCode.json";
857 $data = $this->readJSONFile( $fileName );
858
859 foreach ( $data as $key => $item ) {
860 $this->mergeItem( $key, $allData[$key], $item );
861 }
862
863 $deps[] = new FileDependency( $fileName );
864 }
865 }
866 }
867
868 foreach ( $wgExtensionMessagesFiles as $extension => $fileName ) {
869 $data = $this->readPHPFile( $fileName, 'extension' );
870 $used = false;
871
872 foreach ( $data as $key => $item ) {
873 if ( $key === 'messages' && isset( $wgMessagesDirs[$extension] ) ) {
874 # For backwards compatibility, ignore messages from extensions in
875 # $wgExtensionMessagesFiles that are also present in $wgMessagesDirs.
876 # This allows extensions to use both and be backwards compatible.
877 # Variables other than $messages still need to be supported though.
878 continue;
879 }
880 if ( $this->mergeExtensionItem( $codeSequence, $key, $allData[$key], $item ) ) {
881 $used = true;
882 }
883 }
884
885 if ( $used ) {
886 $deps[] = new FileDependency( $fileName );
887 }
888 }
889
890 # Merge core data into extension data
891 foreach ( $coreData as $key => $item ) {
892 $this->mergeItem( $key, $allData[$key], $item );
893 }
894 wfProfileOut( __METHOD__ . '-extensions' );
895
896 # Add cache dependencies for any referenced globals
897 $deps['wgExtensionMessagesFiles'] = new GlobalDependency( 'wgExtensionMessagesFiles' );
898 $deps['wgMessagesDirs'] = new GlobalDependency( 'wgMessagesDirs' );
899 $deps['version'] = new ConstantDependency( 'MW_LC_VERSION' );
900
901 # Add dependencies to the cache entry
902 $allData['deps'] = $deps;
903
904 # Replace spaces with underscores in namespace names
905 $allData['namespaceNames'] = str_replace( ' ', '_', $allData['namespaceNames'] );
906
907 # And do the same for special page aliases. $page is an array.
908 foreach ( $allData['specialPageAliases'] as &$page ) {
909 $page = str_replace( ' ', '_', $page );
910 }
911 # Decouple the reference to prevent accidental damage
912 unset( $page );
913
914 # If there were no plural rules, return an empty array
915 if ( $allData['pluralRules'] === null ) {
916 $allData['pluralRules'] = array();
917 }
918 if ( $allData['compiledPluralRules'] === null ) {
919 $allData['compiledPluralRules'] = array();
920 }
921 # If there were no plural rule types, return an empty array
922 if ( $allData['pluralRuleTypes'] === null ) {
923 $allData['pluralRuleTypes'] = array();
924 }
925
926 # Set the list keys
927 $allData['list'] = array();
928 foreach ( self::$splitKeys as $key ) {
929 $allData['list'][$key] = array_keys( $allData[$key] );
930 }
931 # Run hooks
932 $purgeBlobs = true;
933 wfRunHooks( 'LocalisationCacheRecache', array( $this, $code, &$allData, &$purgeBlobs ) );
934
935 if ( is_null( $allData['namespaceNames'] ) ) {
936 wfProfileOut( __METHOD__ );
937 throw new MWException( __METHOD__ . ': Localisation data failed sanity check! ' .
938 'Check that your languages/messages/MessagesEn.php file is intact.' );
939 }
940
941 # Set the preload key
942 $allData['preload'] = $this->buildPreload( $allData );
943
944 # Save to the process cache and register the items loaded
945 $this->data[$code] = $allData;
946 foreach ( $allData as $key => $item ) {
947 $this->loadedItems[$code][$key] = true;
948 }
949
950 # Save to the persistent cache
951 wfProfileIn( __METHOD__ . '-write' );
952 $this->store->startWrite( $code );
953 foreach ( $allData as $key => $value ) {
954 if ( in_array( $key, self::$splitKeys ) ) {
955 foreach ( $value as $subkey => $subvalue ) {
956 $this->store->set( "$key:$subkey", $subvalue );
957 }
958 } else {
959 $this->store->set( $key, $value );
960 }
961 }
962 $this->store->finishWrite();
963 wfProfileOut( __METHOD__ . '-write' );
964
965 # Clear out the MessageBlobStore
966 # HACK: If using a null (i.e. disabled) storage backend, we
967 # can't write to the MessageBlobStore either
968 if ( $purgeBlobs && !$this->store instanceof LCStoreNull ) {
969 MessageBlobStore::clear();
970 }
971
972 wfProfileOut( __METHOD__ );
973 }
974
975 /**
976 * Build the preload item from the given pre-cache data.
977 *
978 * The preload item will be loaded automatically, improving performance
979 * for the commonly-requested items it contains.
980 * @param $data
981 * @return array
982 */
983 protected function buildPreload( $data ) {
984 $preload = array( 'messages' => array() );
985 foreach ( self::$preloadedKeys as $key ) {
986 $preload[$key] = $data[$key];
987 }
988
989 foreach ( $data['preloadedMessages'] as $subkey ) {
990 if ( isset( $data['messages'][$subkey] ) ) {
991 $subitem = $data['messages'][$subkey];
992 } else {
993 $subitem = null;
994 }
995 $preload['messages'][$subkey] = $subitem;
996 }
997
998 return $preload;
999 }
1000
1001 /**
1002 * Unload the data for a given language from the object cache.
1003 * Reduces memory usage.
1004 * @param $code
1005 */
1006 public function unload( $code ) {
1007 unset( $this->data[$code] );
1008 unset( $this->loadedItems[$code] );
1009 unset( $this->loadedSubitems[$code] );
1010 unset( $this->initialisedLangs[$code] );
1011 unset( $this->shallowFallbacks[$code] );
1012
1013 foreach ( $this->shallowFallbacks as $shallowCode => $fbCode ) {
1014 if ( $fbCode === $code ) {
1015 $this->unload( $shallowCode );
1016 }
1017 }
1018 }
1019
1020 /**
1021 * Unload all data
1022 */
1023 public function unloadAll() {
1024 foreach ( $this->initialisedLangs as $lang => $unused ) {
1025 $this->unload( $lang );
1026 }
1027 }
1028
1029 /**
1030 * Disable the storage backend
1031 */
1032 public function disableBackend() {
1033 $this->store = new LCStoreNull;
1034 $this->manualRecache = false;
1035 }
1036 }
1037
1038 /**
1039 * Interface for the persistence layer of LocalisationCache.
1040 *
1041 * The persistence layer is two-level hierarchical cache. The first level
1042 * is the language, the second level is the item or subitem.
1043 *
1044 * Since the data for a whole language is rebuilt in one operation, it needs
1045 * to have a fast and atomic method for deleting or replacing all of the
1046 * current data for a given language. The interface reflects this bulk update
1047 * operation. Callers writing to the cache must first call startWrite(), then
1048 * will call set() a couple of thousand times, then will call finishWrite()
1049 * to commit the operation. When finishWrite() is called, the cache is
1050 * expected to delete all data previously stored for that language.
1051 *
1052 * The values stored are PHP variables suitable for serialize(). Implementations
1053 * of LCStore are responsible for serializing and unserializing.
1054 */
1055 interface LCStore {
1056 /**
1057 * Get a value.
1058 * @param string $code Language code
1059 * @param string $key Cache key
1060 */
1061 function get( $code, $key );
1062
1063 /**
1064 * Start a write transaction.
1065 * @param string $code Language code
1066 */
1067 function startWrite( $code );
1068
1069 /**
1070 * Finish a write transaction.
1071 */
1072 function finishWrite();
1073
1074 /**
1075 * Set a key to a given value. startWrite() must be called before this
1076 * is called, and finishWrite() must be called afterwards.
1077 * @param string $key
1078 * @param mixed $value
1079 */
1080 function set( $key, $value );
1081 }
1082
1083 /**
1084 * LCStore implementation which uses PHP accelerator to store data.
1085 * This will work if one of XCache, WinCache or APC cacher is configured.
1086 * (See ObjectCache.php)
1087 */
1088 class LCStoreAccel implements LCStore {
1089 private $currentLang;
1090 private $keys;
1091
1092 public function __construct() {
1093 $this->cache = wfGetCache( CACHE_ACCEL );
1094 }
1095
1096 public function get( $code, $key ) {
1097 $k = wfMemcKey( 'l10n', $code, 'k', $key );
1098 $r = $this->cache->get( $k );
1099
1100 return $r === false ? null : $r;
1101 }
1102
1103 public function startWrite( $code ) {
1104 $k = wfMemcKey( 'l10n', $code, 'l' );
1105 $keys = $this->cache->get( $k );
1106 if ( $keys ) {
1107 foreach ( $keys as $k ) {
1108 $this->cache->delete( $k );
1109 }
1110 }
1111 $this->currentLang = $code;
1112 $this->keys = array();
1113 }
1114
1115 public function finishWrite() {
1116 if ( $this->currentLang ) {
1117 $k = wfMemcKey( 'l10n', $this->currentLang, 'l' );
1118 $this->cache->set( $k, array_keys( $this->keys ) );
1119 }
1120 $this->currentLang = null;
1121 $this->keys = array();
1122 }
1123
1124 public function set( $key, $value ) {
1125 if ( $this->currentLang ) {
1126 $k = wfMemcKey( 'l10n', $this->currentLang, 'k', $key );
1127 $this->keys[$k] = true;
1128 $this->cache->set( $k, $value );
1129 }
1130 }
1131 }
1132
1133 /**
1134 * LCStore implementation which uses the standard DB functions to store data.
1135 * This will work on any MediaWiki installation.
1136 */
1137 class LCStoreDB implements LCStore {
1138 private $currentLang;
1139 private $writesDone = false;
1140
1141 /**
1142 * @var DatabaseBase
1143 */
1144 private $dbw;
1145 private $batch;
1146 private $readOnly = false;
1147
1148 public function get( $code, $key ) {
1149 if ( $this->writesDone ) {
1150 $db = wfGetDB( DB_MASTER );
1151 } else {
1152 $db = wfGetDB( DB_SLAVE );
1153 }
1154 $row = $db->selectRow( 'l10n_cache', array( 'lc_value' ),
1155 array( 'lc_lang' => $code, 'lc_key' => $key ), __METHOD__ );
1156 if ( $row ) {
1157 return unserialize( $row->lc_value );
1158 } else {
1159 return null;
1160 }
1161 }
1162
1163 public function startWrite( $code ) {
1164 if ( $this->readOnly ) {
1165 return;
1166 }
1167
1168 if ( !$code ) {
1169 throw new MWException( __METHOD__ . ": Invalid language \"$code\"" );
1170 }
1171
1172 $this->dbw = wfGetDB( DB_MASTER );
1173 try {
1174 $this->dbw->begin( __METHOD__ );
1175 $this->dbw->delete( 'l10n_cache', array( 'lc_lang' => $code ), __METHOD__ );
1176 } catch ( DBQueryError $e ) {
1177 if ( $this->dbw->wasReadOnlyError() ) {
1178 $this->readOnly = true;
1179 $this->dbw->rollback( __METHOD__ );
1180
1181 return;
1182 } else {
1183 throw $e;
1184 }
1185 }
1186
1187 $this->currentLang = $code;
1188 $this->batch = array();
1189 }
1190
1191 public function finishWrite() {
1192 if ( $this->readOnly ) {
1193 return;
1194 }
1195
1196 if ( $this->batch ) {
1197 $this->dbw->insert( 'l10n_cache', $this->batch, __METHOD__ );
1198 }
1199
1200 $this->dbw->commit( __METHOD__ );
1201 $this->currentLang = null;
1202 $this->dbw = null;
1203 $this->batch = array();
1204 $this->writesDone = true;
1205 }
1206
1207 public function set( $key, $value ) {
1208 if ( $this->readOnly ) {
1209 return;
1210 }
1211
1212 if ( is_null( $this->currentLang ) ) {
1213 throw new MWException( __CLASS__ . ': must call startWrite() before calling set()' );
1214 }
1215
1216 $this->batch[] = array(
1217 'lc_lang' => $this->currentLang,
1218 'lc_key' => $key,
1219 'lc_value' => serialize( $value ) );
1220
1221 if ( count( $this->batch ) >= 100 ) {
1222 $this->dbw->insert( 'l10n_cache', $this->batch, __METHOD__ );
1223 $this->batch = array();
1224 }
1225 }
1226 }
1227
1228 /**
1229 * LCStore implementation which stores data as a collection of CDB files in the
1230 * directory given by $wgCacheDirectory. If $wgCacheDirectory is not set, this
1231 * will throw an exception.
1232 *
1233 * Profiling indicates that on Linux, this implementation outperforms MySQL if
1234 * the directory is on a local filesystem and there is ample kernel cache
1235 * space. The performance advantage is greater when the DBA extension is
1236 * available than it is with the PHP port.
1237 *
1238 * See Cdb.php and http://cr.yp.to/cdb.html
1239 */
1240 class LCStoreCDB implements LCStore {
1241 /** @var CdbReader[] */
1242 private $readers;
1243
1244 /** @var CdbWriter */
1245 private $writer;
1246
1247 /** @var string Current language code */
1248 private $currentLang;
1249
1250 /** @var bool|string Cache directory. False if not set */
1251 private $directory;
1252
1253 function __construct( $conf = array() ) {
1254 global $wgCacheDirectory;
1255
1256 if ( isset( $conf['directory'] ) ) {
1257 $this->directory = $conf['directory'];
1258 } else {
1259 $this->directory = $wgCacheDirectory;
1260 }
1261 }
1262
1263 public function get( $code, $key ) {
1264 if ( !isset( $this->readers[$code] ) ) {
1265 $fileName = $this->getFileName( $code );
1266
1267 $this->readers[$code] = false;
1268 if ( file_exists( $fileName ) ) {
1269 try {
1270 $this->readers[$code] = CdbReader::open( $fileName );
1271 } catch ( CdbException $e ) {
1272 wfDebug( __METHOD__ . ": unable to open cdb file for reading" );
1273 }
1274 }
1275 }
1276
1277 if ( !$this->readers[$code] ) {
1278 return null;
1279 } else {
1280 $value = false;
1281 try {
1282 $value = $this->readers[$code]->get( $key );
1283 } catch ( CdbException $e ) {
1284 wfDebug( __METHOD__ . ": CdbException caught, error message was "
1285 . $e->getMessage() );
1286 }
1287 if ( $value === false ) {
1288 return null;
1289 }
1290
1291 return unserialize( $value );
1292 }
1293 }
1294
1295 public function startWrite( $code ) {
1296 if ( !file_exists( $this->directory ) ) {
1297 if ( !wfMkdirParents( $this->directory, null, __METHOD__ ) ) {
1298 throw new MWException( "Unable to create the localisation store " .
1299 "directory \"{$this->directory}\"" );
1300 }
1301 }
1302
1303 // Close reader to stop permission errors on write
1304 if ( !empty( $this->readers[$code] ) ) {
1305 $this->readers[$code]->close();
1306 }
1307
1308 try {
1309 $this->writer = CdbWriter::open( $this->getFileName( $code ) );
1310 } catch ( CdbException $e ) {
1311 throw new MWException( $e->getMessage() );
1312 }
1313 $this->currentLang = $code;
1314 }
1315
1316 public function finishWrite() {
1317 // Close the writer
1318 try {
1319 $this->writer->close();
1320 } catch ( CdbException $e ) {
1321 throw new MWException( $e->getMessage() );
1322 }
1323 $this->writer = null;
1324 unset( $this->readers[$this->currentLang] );
1325 $this->currentLang = null;
1326 }
1327
1328 public function set( $key, $value ) {
1329 if ( is_null( $this->writer ) ) {
1330 throw new MWException( __CLASS__ . ': must call startWrite() before calling set()' );
1331 }
1332 try {
1333 $this->writer->set( $key, serialize( $value ) );
1334 } catch ( CdbException $e ) {
1335 throw new MWException( $e->getMessage() );
1336 }
1337 }
1338
1339 protected function getFileName( $code ) {
1340 if ( strval( $code ) === '' || strpos( $code, '/' ) !== false ) {
1341 throw new MWException( __METHOD__ . ": Invalid language \"$code\"" );
1342 }
1343
1344 return "{$this->directory}/l10n_cache-$code.cdb";
1345 }
1346 }
1347
1348 /**
1349 * Null store backend, used to avoid DB errors during install
1350 */
1351 class LCStoreNull implements LCStore {
1352 public function get( $code, $key ) {
1353 return null;
1354 }
1355
1356 public function startWrite( $code ) {
1357 }
1358
1359 public function finishWrite() {
1360 }
1361
1362 public function set( $key, $value ) {
1363 }
1364 }
1365
1366 /**
1367 * A localisation cache optimised for loading large amounts of data for many
1368 * languages. Used by rebuildLocalisationCache.php.
1369 */
1370 class LocalisationCacheBulkLoad extends LocalisationCache {
1371 /**
1372 * A cache of the contents of data files.
1373 * Core files are serialized to avoid using ~1GB of RAM during a recache.
1374 */
1375 private $fileCache = array();
1376
1377 /**
1378 * Most recently used languages. Uses the linked-list aspect of PHP hashtables
1379 * to keep the most recently used language codes at the end of the array, and
1380 * the language codes that are ready to be deleted at the beginning.
1381 */
1382 private $mruLangs = array();
1383
1384 /**
1385 * Maximum number of languages that may be loaded into $this->data
1386 */
1387 private $maxLoadedLangs = 10;
1388
1389 /**
1390 * @param $fileName
1391 * @param $fileType
1392 * @return array|mixed
1393 */
1394 protected function readPHPFile( $fileName, $fileType ) {
1395 $serialize = $fileType === 'core';
1396 if ( !isset( $this->fileCache[$fileName][$fileType] ) ) {
1397 $data = parent::readPHPFile( $fileName, $fileType );
1398
1399 if ( $serialize ) {
1400 $encData = serialize( $data );
1401 } else {
1402 $encData = $data;
1403 }
1404
1405 $this->fileCache[$fileName][$fileType] = $encData;
1406
1407 return $data;
1408 } elseif ( $serialize ) {
1409 return unserialize( $this->fileCache[$fileName][$fileType] );
1410 } else {
1411 return $this->fileCache[$fileName][$fileType];
1412 }
1413 }
1414
1415 /**
1416 * @param $code
1417 * @param $key
1418 * @return mixed
1419 */
1420 public function getItem( $code, $key ) {
1421 unset( $this->mruLangs[$code] );
1422 $this->mruLangs[$code] = true;
1423
1424 return parent::getItem( $code, $key );
1425 }
1426
1427 /**
1428 * @param $code
1429 * @param $key
1430 * @param $subkey
1431 * @return
1432 */
1433 public function getSubitem( $code, $key, $subkey ) {
1434 unset( $this->mruLangs[$code] );
1435 $this->mruLangs[$code] = true;
1436
1437 return parent::getSubitem( $code, $key, $subkey );
1438 }
1439
1440 /**
1441 * @param $code
1442 */
1443 public function recache( $code ) {
1444 parent::recache( $code );
1445 unset( $this->mruLangs[$code] );
1446 $this->mruLangs[$code] = true;
1447 $this->trimCache();
1448 }
1449
1450 /**
1451 * @param $code
1452 */
1453 public function unload( $code ) {
1454 unset( $this->mruLangs[$code] );
1455 parent::unload( $code );
1456 }
1457
1458 /**
1459 * Unload cached languages until there are less than $this->maxLoadedLangs
1460 */
1461 protected function trimCache() {
1462 while ( count( $this->data ) > $this->maxLoadedLangs && count( $this->mruLangs ) ) {
1463 reset( $this->mruLangs );
1464 $code = key( $this->mruLangs );
1465 wfDebug( __METHOD__ . ": unloading $code\n" );
1466 $this->unload( $code );
1467 }
1468 }
1469 }