Merge "Add title to list item of language link"
[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
563 // The JSON format only supports messages, none of the other variables, so wrap the data
564 return array( 'messages' => $data );
565 }
566
567 /**
568 * Get the compiled plural rules for a given language from the XML files.
569 * @since 1.20
570 */
571 public function getCompiledPluralRules( $code ) {
572 $rules = $this->getPluralRules( $code );
573 if ( $rules === null ) {
574 return null;
575 }
576 try {
577 $compiledRules = CLDRPluralRuleEvaluator::compile( $rules );
578 } catch ( CLDRPluralRuleError $e ) {
579 wfDebugLog( 'l10n', $e->getMessage() );
580
581 return array();
582 }
583
584 return $compiledRules;
585 }
586
587 /**
588 * Get the plural rules for a given language from the XML files.
589 * Cached.
590 * @since 1.20
591 */
592 public function getPluralRules( $code ) {
593 if ( $this->pluralRules === null ) {
594 $this->loadPluralFiles();
595 }
596 if ( !isset( $this->pluralRules[$code] ) ) {
597 return null;
598 } else {
599 return $this->pluralRules[$code];
600 }
601 }
602
603 /**
604 * Get the plural rule types for a given language from the XML files.
605 * Cached.
606 * @since 1.22
607 */
608 public function getPluralRuleTypes( $code ) {
609 if ( $this->pluralRuleTypes === null ) {
610 $this->loadPluralFiles();
611 }
612 if ( !isset( $this->pluralRuleTypes[$code] ) ) {
613 return null;
614 } else {
615 return $this->pluralRuleTypes[$code];
616 }
617 }
618
619 /**
620 * Load the plural XML files.
621 */
622 protected function loadPluralFiles() {
623 global $IP;
624 $cldrPlural = "$IP/languages/data/plurals.xml";
625 $mwPlural = "$IP/languages/data/plurals-mediawiki.xml";
626 // Load CLDR plural rules
627 $this->loadPluralFile( $cldrPlural );
628 if ( file_exists( $mwPlural ) ) {
629 // Override or extend
630 $this->loadPluralFile( $mwPlural );
631 }
632 }
633
634 /**
635 * Load a plural XML file with the given filename, compile the relevant
636 * rules, and save the compiled rules in a process-local cache.
637 */
638 protected function loadPluralFile( $fileName ) {
639 $doc = new DOMDocument;
640 $doc->load( $fileName );
641 $rulesets = $doc->getElementsByTagName( "pluralRules" );
642 foreach ( $rulesets as $ruleset ) {
643 $codes = $ruleset->getAttribute( 'locales' );
644 $rules = array();
645 $ruleTypes = array();
646 $ruleElements = $ruleset->getElementsByTagName( "pluralRule" );
647 foreach ( $ruleElements as $elt ) {
648 $ruleType = $elt->getAttribute( 'count' );
649 if ( $ruleType === 'other' ) {
650 // Don't record "other" rules, which have an empty condition
651 continue;
652 }
653 $rules[] = $elt->nodeValue;
654 $ruleTypes[] = $ruleType;
655 }
656 foreach ( explode( ' ', $codes ) as $code ) {
657 $this->pluralRules[$code] = $rules;
658 $this->pluralRuleTypes[$code] = $ruleTypes;
659 }
660 }
661 }
662
663 /**
664 * Read the data from the source files for a given language, and register
665 * the relevant dependencies in the $deps array. If the localisation
666 * exists, the data array is returned, otherwise false is returned.
667 */
668 protected function readSourceFilesAndRegisterDeps( $code, &$deps ) {
669 global $IP;
670 wfProfileIn( __METHOD__ );
671
672 $fileName = Language::getMessagesFileName( $code );
673 if ( !file_exists( $fileName ) ) {
674 wfProfileOut( __METHOD__ );
675
676 return false;
677 }
678
679 $deps[] = new FileDependency( $fileName );
680 $data = $this->readPHPFile( $fileName, 'core' );
681
682 # Load CLDR plural rules for JavaScript
683 $data['pluralRules'] = $this->getPluralRules( $code );
684 # And for PHP
685 $data['compiledPluralRules'] = $this->getCompiledPluralRules( $code );
686 # Load plural rule types
687 $data['pluralRuleTypes'] = $this->getPluralRuleTypes( $code );
688
689 $deps['plurals'] = new FileDependency( "$IP/languages/data/plurals.xml" );
690 $deps['plurals-mw'] = new FileDependency( "$IP/languages/data/plurals-mediawiki.xml" );
691
692 wfProfileOut( __METHOD__ );
693
694 return $data;
695 }
696
697 /**
698 * Merge two localisation values, a primary and a fallback, overwriting the
699 * primary value in place.
700 * @param $key
701 * @param $value
702 * @param $fallbackValue
703 */
704 protected function mergeItem( $key, &$value, $fallbackValue ) {
705 if ( !is_null( $value ) ) {
706 if ( !is_null( $fallbackValue ) ) {
707 if ( in_array( $key, self::$mergeableMapKeys ) ) {
708 $value = $value + $fallbackValue;
709 } elseif ( in_array( $key, self::$mergeableListKeys ) ) {
710 $value = array_unique( array_merge( $fallbackValue, $value ) );
711 } elseif ( in_array( $key, self::$mergeableAliasListKeys ) ) {
712 $value = array_merge_recursive( $value, $fallbackValue );
713 } elseif ( in_array( $key, self::$optionalMergeKeys ) ) {
714 if ( !empty( $value['inherit'] ) ) {
715 $value = array_merge( $fallbackValue, $value );
716 }
717
718 if ( isset( $value['inherit'] ) ) {
719 unset( $value['inherit'] );
720 }
721 } elseif ( in_array( $key, self::$magicWordKeys ) ) {
722 $this->mergeMagicWords( $value, $fallbackValue );
723 }
724 }
725 } else {
726 $value = $fallbackValue;
727 }
728 }
729
730 /**
731 * @param $value
732 * @param $fallbackValue
733 */
734 protected function mergeMagicWords( &$value, $fallbackValue ) {
735 foreach ( $fallbackValue as $magicName => $fallbackInfo ) {
736 if ( !isset( $value[$magicName] ) ) {
737 $value[$magicName] = $fallbackInfo;
738 } else {
739 $oldSynonyms = array_slice( $fallbackInfo, 1 );
740 $newSynonyms = array_slice( $value[$magicName], 1 );
741 $synonyms = array_values( array_unique( array_merge(
742 $newSynonyms, $oldSynonyms ) ) );
743 $value[$magicName] = array_merge( array( $fallbackInfo[0] ), $synonyms );
744 }
745 }
746 }
747
748 /**
749 * Given an array mapping language code to localisation value, such as is
750 * found in extension *.i18n.php files, iterate through a fallback sequence
751 * to merge the given data with an existing primary value.
752 *
753 * Returns true if any data from the extension array was used, false
754 * otherwise.
755 * @param $codeSequence
756 * @param $key
757 * @param $value
758 * @param $fallbackValue
759 * @return bool
760 */
761 protected function mergeExtensionItem( $codeSequence, $key, &$value, $fallbackValue ) {
762 $used = false;
763 foreach ( $codeSequence as $code ) {
764 if ( isset( $fallbackValue[$code] ) ) {
765 $this->mergeItem( $key, $value, $fallbackValue[$code] );
766 $used = true;
767 }
768 }
769
770 return $used;
771 }
772
773 /**
774 * Load localisation data for a given language for both core and extensions
775 * and save it to the persistent cache store and the process cache
776 * @param $code
777 * @throws MWException
778 */
779 public function recache( $code ) {
780 global $wgExtensionMessagesFiles, $wgMessagesDirs;
781 wfProfileIn( __METHOD__ );
782
783 if ( !$code ) {
784 wfProfileOut( __METHOD__ );
785 throw new MWException( "Invalid language code requested" );
786 }
787 $this->recachedLangs[$code] = true;
788
789 # Initial values
790 $initialData = array_combine(
791 self::$allKeys,
792 array_fill( 0, count( self::$allKeys ), null ) );
793 $coreData = $initialData;
794 $deps = array();
795
796 # Load the primary localisation from the source file
797 $data = $this->readSourceFilesAndRegisterDeps( $code, $deps );
798 if ( $data === false ) {
799 wfDebug( __METHOD__ . ": no localisation file for $code, using fallback to en\n" );
800 $coreData['fallback'] = 'en';
801 } else {
802 wfDebug( __METHOD__ . ": got localisation for $code from source\n" );
803
804 # Merge primary localisation
805 foreach ( $data as $key => $value ) {
806 $this->mergeItem( $key, $coreData[$key], $value );
807 }
808 }
809
810 # Fill in the fallback if it's not there already
811 if ( is_null( $coreData['fallback'] ) ) {
812 $coreData['fallback'] = $code === 'en' ? false : 'en';
813 }
814 if ( $coreData['fallback'] === false ) {
815 $coreData['fallbackSequence'] = array();
816 } else {
817 $coreData['fallbackSequence'] = array_map( 'trim', explode( ',', $coreData['fallback'] ) );
818 $len = count( $coreData['fallbackSequence'] );
819
820 # Ensure that the sequence ends at en
821 if ( $coreData['fallbackSequence'][$len - 1] !== 'en' ) {
822 $coreData['fallbackSequence'][] = 'en';
823 }
824
825 # Load the fallback localisation item by item and merge it
826 foreach ( $coreData['fallbackSequence'] as $fbCode ) {
827 # Load the secondary localisation from the source file to
828 # avoid infinite cycles on cyclic fallbacks
829 $fbData = $this->readSourceFilesAndRegisterDeps( $fbCode, $deps );
830 if ( $fbData === false ) {
831 continue;
832 }
833
834 foreach ( self::$allKeys as $key ) {
835 if ( !isset( $fbData[$key] ) ) {
836 continue;
837 }
838
839 if ( is_null( $coreData[$key] ) || $this->isMergeableKey( $key ) ) {
840 $this->mergeItem( $key, $coreData[$key], $fbData[$key] );
841 }
842 }
843 }
844 }
845
846 $codeSequence = array_merge( array( $code ), $coreData['fallbackSequence'] );
847
848 # Load the extension localisations
849 # This is done after the core because we know the fallback sequence now.
850 # But it has a higher precedence for merging so that we can support things
851 # like site-specific message overrides.
852 wfProfileIn( __METHOD__ . '-extensions' );
853 $allData = $initialData;
854 foreach ( $wgMessagesDirs as $dirs ) {
855 foreach ( (array)$dirs as $dir ) {
856 foreach ( $codeSequence as $csCode ) {
857 $fileName = "$dir/$csCode.json";
858 $data = $this->readJSONFile( $fileName );
859
860 foreach ( $data as $key => $item ) {
861 $this->mergeItem( $key, $allData[$key], $item );
862 }
863
864 $deps[] = new FileDependency( $fileName );
865 }
866 }
867 }
868
869 foreach ( $wgExtensionMessagesFiles as $extension => $fileName ) {
870 $data = $this->readPHPFile( $fileName, 'extension' );
871 $used = false;
872
873 foreach ( $data as $key => $item ) {
874 if ( $key === 'messages' && isset( $wgMessagesDirs[$extension] ) ) {
875 # For backwards compatibility, ignore messages from extensions in
876 # $wgExtensionMessagesFiles that are also present in $wgMessagesDirs.
877 # This allows extensions to use both and be backwards compatible.
878 # Variables other than $messages still need to be supported though.
879 continue;
880 }
881 if ( $this->mergeExtensionItem( $codeSequence, $key, $allData[$key], $item ) ) {
882 $used = true;
883 }
884 }
885
886 if ( $used ) {
887 $deps[] = new FileDependency( $fileName );
888 }
889 }
890
891 # Merge core data into extension data
892 foreach ( $coreData as $key => $item ) {
893 $this->mergeItem( $key, $allData[$key], $item );
894 }
895 wfProfileOut( __METHOD__ . '-extensions' );
896
897 # Add cache dependencies for any referenced globals
898 $deps['wgExtensionMessagesFiles'] = new GlobalDependency( 'wgExtensionMessagesFiles' );
899 $deps['wgMessagesDirs'] = new GlobalDependency( 'wgMessagesDirs' );
900 $deps['version'] = new ConstantDependency( 'MW_LC_VERSION' );
901
902 # Add dependencies to the cache entry
903 $allData['deps'] = $deps;
904
905 # Replace spaces with underscores in namespace names
906 $allData['namespaceNames'] = str_replace( ' ', '_', $allData['namespaceNames'] );
907
908 # And do the same for special page aliases. $page is an array.
909 foreach ( $allData['specialPageAliases'] as &$page ) {
910 $page = str_replace( ' ', '_', $page );
911 }
912 # Decouple the reference to prevent accidental damage
913 unset( $page );
914
915 # If there were no plural rules, return an empty array
916 if ( $allData['pluralRules'] === null ) {
917 $allData['pluralRules'] = array();
918 }
919 if ( $allData['compiledPluralRules'] === null ) {
920 $allData['compiledPluralRules'] = array();
921 }
922 # If there were no plural rule types, return an empty array
923 if ( $allData['pluralRuleTypes'] === null ) {
924 $allData['pluralRuleTypes'] = array();
925 }
926
927 # Set the list keys
928 $allData['list'] = array();
929 foreach ( self::$splitKeys as $key ) {
930 $allData['list'][$key] = array_keys( $allData[$key] );
931 }
932 # Run hooks
933 $purgeBlobs = true;
934 wfRunHooks( 'LocalisationCacheRecache', array( $this, $code, &$allData, &$purgeBlobs ) );
935
936 if ( is_null( $allData['namespaceNames'] ) ) {
937 wfProfileOut( __METHOD__ );
938 throw new MWException( __METHOD__ . ': Localisation data failed sanity check! ' .
939 'Check that your languages/messages/MessagesEn.php file is intact.' );
940 }
941
942 # Set the preload key
943 $allData['preload'] = $this->buildPreload( $allData );
944
945 # Save to the process cache and register the items loaded
946 $this->data[$code] = $allData;
947 foreach ( $allData as $key => $item ) {
948 $this->loadedItems[$code][$key] = true;
949 }
950
951 # Save to the persistent cache
952 wfProfileIn( __METHOD__ . '-write' );
953 $this->store->startWrite( $code );
954 foreach ( $allData as $key => $value ) {
955 if ( in_array( $key, self::$splitKeys ) ) {
956 foreach ( $value as $subkey => $subvalue ) {
957 $this->store->set( "$key:$subkey", $subvalue );
958 }
959 } else {
960 $this->store->set( $key, $value );
961 }
962 }
963 $this->store->finishWrite();
964 wfProfileOut( __METHOD__ . '-write' );
965
966 # Clear out the MessageBlobStore
967 # HACK: If using a null (i.e. disabled) storage backend, we
968 # can't write to the MessageBlobStore either
969 if ( $purgeBlobs && !$this->store instanceof LCStoreNull ) {
970 MessageBlobStore::clear();
971 }
972
973 wfProfileOut( __METHOD__ );
974 }
975
976 /**
977 * Build the preload item from the given pre-cache data.
978 *
979 * The preload item will be loaded automatically, improving performance
980 * for the commonly-requested items it contains.
981 * @param $data
982 * @return array
983 */
984 protected function buildPreload( $data ) {
985 $preload = array( 'messages' => array() );
986 foreach ( self::$preloadedKeys as $key ) {
987 $preload[$key] = $data[$key];
988 }
989
990 foreach ( $data['preloadedMessages'] as $subkey ) {
991 if ( isset( $data['messages'][$subkey] ) ) {
992 $subitem = $data['messages'][$subkey];
993 } else {
994 $subitem = null;
995 }
996 $preload['messages'][$subkey] = $subitem;
997 }
998
999 return $preload;
1000 }
1001
1002 /**
1003 * Unload the data for a given language from the object cache.
1004 * Reduces memory usage.
1005 * @param $code
1006 */
1007 public function unload( $code ) {
1008 unset( $this->data[$code] );
1009 unset( $this->loadedItems[$code] );
1010 unset( $this->loadedSubitems[$code] );
1011 unset( $this->initialisedLangs[$code] );
1012 unset( $this->shallowFallbacks[$code] );
1013
1014 foreach ( $this->shallowFallbacks as $shallowCode => $fbCode ) {
1015 if ( $fbCode === $code ) {
1016 $this->unload( $shallowCode );
1017 }
1018 }
1019 }
1020
1021 /**
1022 * Unload all data
1023 */
1024 public function unloadAll() {
1025 foreach ( $this->initialisedLangs as $lang => $unused ) {
1026 $this->unload( $lang );
1027 }
1028 }
1029
1030 /**
1031 * Disable the storage backend
1032 */
1033 public function disableBackend() {
1034 $this->store = new LCStoreNull;
1035 $this->manualRecache = false;
1036 }
1037 }
1038
1039 /**
1040 * Interface for the persistence layer of LocalisationCache.
1041 *
1042 * The persistence layer is two-level hierarchical cache. The first level
1043 * is the language, the second level is the item or subitem.
1044 *
1045 * Since the data for a whole language is rebuilt in one operation, it needs
1046 * to have a fast and atomic method for deleting or replacing all of the
1047 * current data for a given language. The interface reflects this bulk update
1048 * operation. Callers writing to the cache must first call startWrite(), then
1049 * will call set() a couple of thousand times, then will call finishWrite()
1050 * to commit the operation. When finishWrite() is called, the cache is
1051 * expected to delete all data previously stored for that language.
1052 *
1053 * The values stored are PHP variables suitable for serialize(). Implementations
1054 * of LCStore are responsible for serializing and unserializing.
1055 */
1056 interface LCStore {
1057 /**
1058 * Get a value.
1059 * @param string $code Language code
1060 * @param string $key Cache key
1061 */
1062 function get( $code, $key );
1063
1064 /**
1065 * Start a write transaction.
1066 * @param string $code Language code
1067 */
1068 function startWrite( $code );
1069
1070 /**
1071 * Finish a write transaction.
1072 */
1073 function finishWrite();
1074
1075 /**
1076 * Set a key to a given value. startWrite() must be called before this
1077 * is called, and finishWrite() must be called afterwards.
1078 * @param string $key
1079 * @param mixed $value
1080 */
1081 function set( $key, $value );
1082 }
1083
1084 /**
1085 * LCStore implementation which uses PHP accelerator to store data.
1086 * This will work if one of XCache, WinCache or APC cacher is configured.
1087 * (See ObjectCache.php)
1088 */
1089 class LCStoreAccel implements LCStore {
1090 private $currentLang;
1091 private $keys;
1092
1093 public function __construct() {
1094 $this->cache = wfGetCache( CACHE_ACCEL );
1095 }
1096
1097 public function get( $code, $key ) {
1098 $k = wfMemcKey( 'l10n', $code, 'k', $key );
1099 $r = $this->cache->get( $k );
1100
1101 return $r === false ? null : $r;
1102 }
1103
1104 public function startWrite( $code ) {
1105 $k = wfMemcKey( 'l10n', $code, 'l' );
1106 $keys = $this->cache->get( $k );
1107 if ( $keys ) {
1108 foreach ( $keys as $k ) {
1109 $this->cache->delete( $k );
1110 }
1111 }
1112 $this->currentLang = $code;
1113 $this->keys = array();
1114 }
1115
1116 public function finishWrite() {
1117 if ( $this->currentLang ) {
1118 $k = wfMemcKey( 'l10n', $this->currentLang, 'l' );
1119 $this->cache->set( $k, array_keys( $this->keys ) );
1120 }
1121 $this->currentLang = null;
1122 $this->keys = array();
1123 }
1124
1125 public function set( $key, $value ) {
1126 if ( $this->currentLang ) {
1127 $k = wfMemcKey( 'l10n', $this->currentLang, 'k', $key );
1128 $this->keys[$k] = true;
1129 $this->cache->set( $k, $value );
1130 }
1131 }
1132 }
1133
1134 /**
1135 * LCStore implementation which uses the standard DB functions to store data.
1136 * This will work on any MediaWiki installation.
1137 */
1138 class LCStoreDB implements LCStore {
1139 private $currentLang;
1140 private $writesDone = false;
1141
1142 /**
1143 * @var DatabaseBase
1144 */
1145 private $dbw;
1146 private $batch;
1147 private $readOnly = false;
1148
1149 public function get( $code, $key ) {
1150 if ( $this->writesDone ) {
1151 $db = wfGetDB( DB_MASTER );
1152 } else {
1153 $db = wfGetDB( DB_SLAVE );
1154 }
1155 $row = $db->selectRow( 'l10n_cache', array( 'lc_value' ),
1156 array( 'lc_lang' => $code, 'lc_key' => $key ), __METHOD__ );
1157 if ( $row ) {
1158 return unserialize( $row->lc_value );
1159 } else {
1160 return null;
1161 }
1162 }
1163
1164 public function startWrite( $code ) {
1165 if ( $this->readOnly ) {
1166 return;
1167 }
1168
1169 if ( !$code ) {
1170 throw new MWException( __METHOD__ . ": Invalid language \"$code\"" );
1171 }
1172
1173 $this->dbw = wfGetDB( DB_MASTER );
1174 try {
1175 $this->dbw->begin( __METHOD__ );
1176 $this->dbw->delete( 'l10n_cache', array( 'lc_lang' => $code ), __METHOD__ );
1177 } catch ( DBQueryError $e ) {
1178 if ( $this->dbw->wasReadOnlyError() ) {
1179 $this->readOnly = true;
1180 $this->dbw->rollback( __METHOD__ );
1181
1182 return;
1183 } else {
1184 throw $e;
1185 }
1186 }
1187
1188 $this->currentLang = $code;
1189 $this->batch = array();
1190 }
1191
1192 public function finishWrite() {
1193 if ( $this->readOnly ) {
1194 return;
1195 }
1196
1197 if ( $this->batch ) {
1198 $this->dbw->insert( 'l10n_cache', $this->batch, __METHOD__ );
1199 }
1200
1201 $this->dbw->commit( __METHOD__ );
1202 $this->currentLang = null;
1203 $this->dbw = null;
1204 $this->batch = array();
1205 $this->writesDone = true;
1206 }
1207
1208 public function set( $key, $value ) {
1209 if ( $this->readOnly ) {
1210 return;
1211 }
1212
1213 if ( is_null( $this->currentLang ) ) {
1214 throw new MWException( __CLASS__ . ': must call startWrite() before calling set()' );
1215 }
1216
1217 $this->batch[] = array(
1218 'lc_lang' => $this->currentLang,
1219 'lc_key' => $key,
1220 'lc_value' => serialize( $value ) );
1221
1222 if ( count( $this->batch ) >= 100 ) {
1223 $this->dbw->insert( 'l10n_cache', $this->batch, __METHOD__ );
1224 $this->batch = array();
1225 }
1226 }
1227 }
1228
1229 /**
1230 * LCStore implementation which stores data as a collection of CDB files in the
1231 * directory given by $wgCacheDirectory. If $wgCacheDirectory is not set, this
1232 * will throw an exception.
1233 *
1234 * Profiling indicates that on Linux, this implementation outperforms MySQL if
1235 * the directory is on a local filesystem and there is ample kernel cache
1236 * space. The performance advantage is greater when the DBA extension is
1237 * available than it is with the PHP port.
1238 *
1239 * See Cdb.php and http://cr.yp.to/cdb.html
1240 */
1241 class LCStoreCDB implements LCStore {
1242 /** @var CdbReader[] */
1243 private $readers;
1244
1245 /** @var CdbWriter */
1246 private $writer;
1247
1248 /** @var string Current language code */
1249 private $currentLang;
1250
1251 /** @var bool|string Cache directory. False if not set */
1252 private $directory;
1253
1254 function __construct( $conf = array() ) {
1255 global $wgCacheDirectory;
1256
1257 if ( isset( $conf['directory'] ) ) {
1258 $this->directory = $conf['directory'];
1259 } else {
1260 $this->directory = $wgCacheDirectory;
1261 }
1262 }
1263
1264 public function get( $code, $key ) {
1265 if ( !isset( $this->readers[$code] ) ) {
1266 $fileName = $this->getFileName( $code );
1267
1268 $this->readers[$code] = false;
1269 if ( file_exists( $fileName ) ) {
1270 try {
1271 $this->readers[$code] = CdbReader::open( $fileName );
1272 } catch ( CdbException $e ) {
1273 wfDebug( __METHOD__ . ": unable to open cdb file for reading" );
1274 }
1275 }
1276 }
1277
1278 if ( !$this->readers[$code] ) {
1279 return null;
1280 } else {
1281 $value = false;
1282 try {
1283 $value = $this->readers[$code]->get( $key );
1284 } catch ( CdbException $e ) {
1285 wfDebug( __METHOD__ . ": CdbException caught, error message was "
1286 . $e->getMessage() );
1287 }
1288 if ( $value === false ) {
1289 return null;
1290 }
1291
1292 return unserialize( $value );
1293 }
1294 }
1295
1296 public function startWrite( $code ) {
1297 if ( !file_exists( $this->directory ) ) {
1298 if ( !wfMkdirParents( $this->directory, null, __METHOD__ ) ) {
1299 throw new MWException( "Unable to create the localisation store " .
1300 "directory \"{$this->directory}\"" );
1301 }
1302 }
1303
1304 // Close reader to stop permission errors on write
1305 if ( !empty( $this->readers[$code] ) ) {
1306 $this->readers[$code]->close();
1307 }
1308
1309 try {
1310 $this->writer = CdbWriter::open( $this->getFileName( $code ) );
1311 } catch ( CdbException $e ) {
1312 throw new MWException( $e->getMessage() );
1313 }
1314 $this->currentLang = $code;
1315 }
1316
1317 public function finishWrite() {
1318 // Close the writer
1319 try {
1320 $this->writer->close();
1321 } catch ( CdbException $e ) {
1322 throw new MWException( $e->getMessage() );
1323 }
1324 $this->writer = null;
1325 unset( $this->readers[$this->currentLang] );
1326 $this->currentLang = null;
1327 }
1328
1329 public function set( $key, $value ) {
1330 if ( is_null( $this->writer ) ) {
1331 throw new MWException( __CLASS__ . ': must call startWrite() before calling set()' );
1332 }
1333 try {
1334 $this->writer->set( $key, serialize( $value ) );
1335 } catch ( CdbException $e ) {
1336 throw new MWException( $e->getMessage() );
1337 }
1338 }
1339
1340 protected function getFileName( $code ) {
1341 if ( strval( $code ) === '' || strpos( $code, '/' ) !== false ) {
1342 throw new MWException( __METHOD__ . ": Invalid language \"$code\"" );
1343 }
1344
1345 return "{$this->directory}/l10n_cache-$code.cdb";
1346 }
1347 }
1348
1349 /**
1350 * Null store backend, used to avoid DB errors during install
1351 */
1352 class LCStoreNull implements LCStore {
1353 public function get( $code, $key ) {
1354 return null;
1355 }
1356
1357 public function startWrite( $code ) {
1358 }
1359
1360 public function finishWrite() {
1361 }
1362
1363 public function set( $key, $value ) {
1364 }
1365 }
1366
1367 /**
1368 * A localisation cache optimised for loading large amounts of data for many
1369 * languages. Used by rebuildLocalisationCache.php.
1370 */
1371 class LocalisationCacheBulkLoad extends LocalisationCache {
1372 /**
1373 * A cache of the contents of data files.
1374 * Core files are serialized to avoid using ~1GB of RAM during a recache.
1375 */
1376 private $fileCache = array();
1377
1378 /**
1379 * Most recently used languages. Uses the linked-list aspect of PHP hashtables
1380 * to keep the most recently used language codes at the end of the array, and
1381 * the language codes that are ready to be deleted at the beginning.
1382 */
1383 private $mruLangs = array();
1384
1385 /**
1386 * Maximum number of languages that may be loaded into $this->data
1387 */
1388 private $maxLoadedLangs = 10;
1389
1390 /**
1391 * @param $fileName
1392 * @param $fileType
1393 * @return array|mixed
1394 */
1395 protected function readPHPFile( $fileName, $fileType ) {
1396 $serialize = $fileType === 'core';
1397 if ( !isset( $this->fileCache[$fileName][$fileType] ) ) {
1398 $data = parent::readPHPFile( $fileName, $fileType );
1399
1400 if ( $serialize ) {
1401 $encData = serialize( $data );
1402 } else {
1403 $encData = $data;
1404 }
1405
1406 $this->fileCache[$fileName][$fileType] = $encData;
1407
1408 return $data;
1409 } elseif ( $serialize ) {
1410 return unserialize( $this->fileCache[$fileName][$fileType] );
1411 } else {
1412 return $this->fileCache[$fileName][$fileType];
1413 }
1414 }
1415
1416 /**
1417 * @param $code
1418 * @param $key
1419 * @return mixed
1420 */
1421 public function getItem( $code, $key ) {
1422 unset( $this->mruLangs[$code] );
1423 $this->mruLangs[$code] = true;
1424
1425 return parent::getItem( $code, $key );
1426 }
1427
1428 /**
1429 * @param $code
1430 * @param $key
1431 * @param $subkey
1432 * @return
1433 */
1434 public function getSubitem( $code, $key, $subkey ) {
1435 unset( $this->mruLangs[$code] );
1436 $this->mruLangs[$code] = true;
1437
1438 return parent::getSubitem( $code, $key, $subkey );
1439 }
1440
1441 /**
1442 * @param $code
1443 */
1444 public function recache( $code ) {
1445 parent::recache( $code );
1446 unset( $this->mruLangs[$code] );
1447 $this->mruLangs[$code] = true;
1448 $this->trimCache();
1449 }
1450
1451 /**
1452 * @param $code
1453 */
1454 public function unload( $code ) {
1455 unset( $this->mruLangs[$code] );
1456 parent::unload( $code );
1457 }
1458
1459 /**
1460 * Unload cached languages until there are less than $this->maxLoadedLangs
1461 */
1462 protected function trimCache() {
1463 while ( count( $this->data ) > $this->maxLoadedLangs && count( $this->mruLangs ) ) {
1464 reset( $this->mruLangs );
1465 $code = key( $this->mruLangs );
1466 wfDebug( __METHOD__ . ": unloading $code\n" );
1467 $this->unload( $code );
1468 }
1469 }
1470 }