Drop outdated "documentation reviewed" tags
[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 use Cdb\Exception as CdbException;
24 use Cdb\Reader as CdbReader;
25 use Cdb\Writer as CdbWriter;
26
27 /**
28 * Class for caching the contents of localisation files, Messages*.php
29 * and *.i18n.php.
30 *
31 * An instance of this class is available using Language::getLocalisationCache().
32 *
33 * The values retrieved from here are merged, containing items from extension
34 * files, core messages files and the language fallback sequence (e.g. zh-cn ->
35 * zh-hans -> en ). Some common errors are corrected, for example namespace
36 * names with spaces instead of underscores, but heavyweight processing, such
37 * as grammatical transformation, is done by the caller.
38 */
39 class LocalisationCache {
40 const VERSION = 3;
41
42 /** Configuration associative array */
43 private $conf;
44
45 /**
46 * True if recaching should only be done on an explicit call to recache().
47 * Setting this reduces the overhead of cache freshness checking, which
48 * requires doing a stat() for every extension i18n file.
49 */
50 private $manualRecache = false;
51
52 /**
53 * True to treat all files as expired until they are regenerated by this object.
54 */
55 private $forceRecache = false;
56
57 /**
58 * The cache data. 3-d array, where the first key is the language code,
59 * the second key is the item key e.g. 'messages', and the third key is
60 * an item specific subkey index. Some items are not arrays and so for those
61 * items, there are no subkeys.
62 */
63 protected $data = array();
64
65 /**
66 * The persistent store object. An instance of LCStore.
67 *
68 * @var LCStore
69 */
70 private $store;
71
72 /**
73 * A 2-d associative array, code/key, where presence indicates that the item
74 * is loaded. Value arbitrary.
75 *
76 * For split items, if set, this indicates that all of the subitems have been
77 * loaded.
78 */
79 private $loadedItems = array();
80
81 /**
82 * A 3-d associative array, code/key/subkey, where presence indicates that
83 * the subitem is loaded. Only used for the split items, i.e. messages.
84 */
85 private $loadedSubitems = array();
86
87 /**
88 * An array where presence of a key indicates that that language has been
89 * initialised. Initialisation includes checking for cache expiry and doing
90 * any necessary updates.
91 */
92 private $initialisedLangs = array();
93
94 /**
95 * An array mapping non-existent pseudo-languages to fallback languages. This
96 * is filled by initShallowFallback() when data is requested from a language
97 * that lacks a Messages*.php file.
98 */
99 private $shallowFallbacks = array();
100
101 /**
102 * An array where the keys are codes that have been recached by this instance.
103 */
104 private $recachedLangs = array();
105
106 /**
107 * All item keys
108 */
109 static public $allKeys = array(
110 'fallback', 'namespaceNames', 'bookstoreList',
111 'magicWords', 'messages', 'rtl', 'capitalizeAllNouns', 'digitTransformTable',
112 'separatorTransformTable', 'fallback8bitEncoding', 'linkPrefixExtension',
113 'linkTrail', 'linkPrefixCharset', 'namespaceAliases',
114 'dateFormats', 'datePreferences', 'datePreferenceMigrationMap',
115 'defaultDateFormat', 'extraUserToggles', 'specialPageAliases',
116 'imageFiles', 'preloadedMessages', 'namespaceGenderAliases',
117 'digitGroupingPattern', 'pluralRules', 'pluralRuleTypes', 'compiledPluralRules',
118 );
119
120 /**
121 * Keys for items which consist of associative arrays, which may be merged
122 * by a fallback sequence.
123 */
124 static public $mergeableMapKeys = array( 'messages', 'namespaceNames',
125 'dateFormats', 'imageFiles', 'preloadedMessages'
126 );
127
128 /**
129 * Keys for items which are a numbered array.
130 */
131 static public $mergeableListKeys = array( 'extraUserToggles' );
132
133 /**
134 * Keys for items which contain an array of arrays of equivalent aliases
135 * for each subitem. The aliases may be merged by a fallback sequence.
136 */
137 static public $mergeableAliasListKeys = array( 'specialPageAliases' );
138
139 /**
140 * Keys for items which contain an associative array, and may be merged if
141 * the primary value contains the special array key "inherit". That array
142 * key is removed after the first merge.
143 */
144 static public $optionalMergeKeys = array( 'bookstoreList' );
145
146 /**
147 * Keys for items that are formatted like $magicWords
148 */
149 static public $magicWordKeys = array( 'magicWords' );
150
151 /**
152 * Keys for items where the subitems are stored in the backend separately.
153 */
154 static public $splitKeys = array( 'messages' );
155
156 /**
157 * Keys which are loaded automatically by initLanguage()
158 */
159 static public $preloadedKeys = array( 'dateFormats', 'namespaceNames' );
160
161 /**
162 * Associative array of cached plural rules. The key is the language code,
163 * the value is an array of plural rules for that language.
164 */
165 private $pluralRules = null;
166
167 /**
168 * Associative array of cached plural rule types. The key is the language
169 * code, the value is an array of plural rule types for that language. For
170 * example, $pluralRuleTypes['ar'] = ['zero', 'one', 'two', 'few', 'many'].
171 * The index for each rule type matches the index for the rule in
172 * $pluralRules, thus allowing correlation between the two. The reason we
173 * don't just use the type names as the keys in $pluralRules is because
174 * Language::convertPlural applies the rules based on numeric order (or
175 * explicit numeric parameter), not based on the name of the rule type. For
176 * example, {{plural:count|wordform1|wordform2|wordform3}}, rather than
177 * {{plural:count|one=wordform1|two=wordform2|many=wordform3}}.
178 */
179 private $pluralRuleTypes = null;
180
181 private $mergeableKeys = null;
182
183 /**
184 * Constructor.
185 * For constructor parameters, see the documentation in DefaultSettings.php
186 * for $wgLocalisationCacheConf.
187 *
188 * @param array $conf
189 * @throws MWException
190 */
191 function __construct( $conf ) {
192 global $wgCacheDirectory;
193
194 $this->conf = $conf;
195 $storeConf = array();
196 if ( !empty( $conf['storeClass'] ) ) {
197 $storeClass = $conf['storeClass'];
198 } else {
199 switch ( $conf['store'] ) {
200 case 'files':
201 case 'file':
202 $storeClass = 'LCStoreCDB';
203 break;
204 case 'db':
205 $storeClass = 'LCStoreDB';
206 break;
207 case 'detect':
208 $storeClass = $wgCacheDirectory ? 'LCStoreCDB' : 'LCStoreDB';
209 break;
210 default:
211 throw new MWException(
212 'Please set $wgLocalisationCacheConf[\'store\'] to something sensible.' );
213 }
214 }
215
216 wfDebugLog( 'caches', get_class( $this ) . ": using store $storeClass" );
217 if ( !empty( $conf['storeDirectory'] ) ) {
218 $storeConf['directory'] = $conf['storeDirectory'];
219 }
220
221 $this->store = new $storeClass( $storeConf );
222 foreach ( array( 'manualRecache', 'forceRecache' ) as $var ) {
223 if ( isset( $conf[$var] ) ) {
224 $this->$var = $conf[$var];
225 }
226 }
227 }
228
229 /**
230 * Returns true if the given key is mergeable, that is, if it is an associative
231 * array which can be merged through a fallback sequence.
232 * @param string $key
233 * @return bool
234 */
235 public function isMergeableKey( $key ) {
236 if ( $this->mergeableKeys === null ) {
237 $this->mergeableKeys = array_flip( array_merge(
238 self::$mergeableMapKeys,
239 self::$mergeableListKeys,
240 self::$mergeableAliasListKeys,
241 self::$optionalMergeKeys,
242 self::$magicWordKeys
243 ) );
244 }
245
246 return isset( $this->mergeableKeys[$key] );
247 }
248
249 /**
250 * Get a cache item.
251 *
252 * Warning: this may be slow for split items (messages), since it will
253 * need to fetch all of the subitems from the cache individually.
254 * @param string $code
255 * @param string $key
256 * @return mixed
257 */
258 public function getItem( $code, $key ) {
259 if ( !isset( $this->loadedItems[$code][$key] ) ) {
260 wfProfileIn( __METHOD__ . '-load' );
261 $this->loadItem( $code, $key );
262 wfProfileOut( __METHOD__ . '-load' );
263 }
264
265 if ( $key === 'fallback' && isset( $this->shallowFallbacks[$code] ) ) {
266 return $this->shallowFallbacks[$code];
267 }
268
269 return $this->data[$code][$key];
270 }
271
272 /**
273 * Get a subitem, for instance a single message for a given language.
274 * @param string $code
275 * @param string $key
276 * @param string $subkey
277 * @return mixed|null
278 */
279 public function getSubitem( $code, $key, $subkey ) {
280 if ( !isset( $this->loadedSubitems[$code][$key][$subkey] ) &&
281 !isset( $this->loadedItems[$code][$key] )
282 ) {
283 wfProfileIn( __METHOD__ . '-load' );
284 $this->loadSubitem( $code, $key, $subkey );
285 wfProfileOut( __METHOD__ . '-load' );
286 }
287
288 if ( isset( $this->data[$code][$key][$subkey] ) ) {
289 return $this->data[$code][$key][$subkey];
290 } else {
291 return null;
292 }
293 }
294
295 /**
296 * Get the list of subitem keys for a given item.
297 *
298 * This is faster than array_keys($lc->getItem(...)) for the items listed in
299 * self::$splitKeys.
300 *
301 * Will return null if the item is not found, or false if the item is not an
302 * array.
303 * @param string $code
304 * @param string $key
305 * @return bool|null|string
306 */
307 public function getSubitemList( $code, $key ) {
308 if ( in_array( $key, self::$splitKeys ) ) {
309 return $this->getSubitem( $code, 'list', $key );
310 } else {
311 $item = $this->getItem( $code, $key );
312 if ( is_array( $item ) ) {
313 return array_keys( $item );
314 } else {
315 return false;
316 }
317 }
318 }
319
320 /**
321 * Load an item into the cache.
322 * @param string $code
323 * @param string $key
324 */
325 protected function loadItem( $code, $key ) {
326 if ( !isset( $this->initialisedLangs[$code] ) ) {
327 $this->initLanguage( $code );
328 }
329
330 // Check to see if initLanguage() loaded it for us
331 if ( isset( $this->loadedItems[$code][$key] ) ) {
332 return;
333 }
334
335 if ( isset( $this->shallowFallbacks[$code] ) ) {
336 $this->loadItem( $this->shallowFallbacks[$code], $key );
337
338 return;
339 }
340
341 if ( in_array( $key, self::$splitKeys ) ) {
342 $subkeyList = $this->getSubitem( $code, 'list', $key );
343 foreach ( $subkeyList as $subkey ) {
344 if ( isset( $this->data[$code][$key][$subkey] ) ) {
345 continue;
346 }
347 $this->data[$code][$key][$subkey] = $this->getSubitem( $code, $key, $subkey );
348 }
349 } else {
350 $this->data[$code][$key] = $this->store->get( $code, $key );
351 }
352
353 $this->loadedItems[$code][$key] = true;
354 }
355
356 /**
357 * Load a subitem into the cache
358 * @param string $code
359 * @param string $key
360 * @param string $subkey
361 */
362 protected function loadSubitem( $code, $key, $subkey ) {
363 if ( !in_array( $key, self::$splitKeys ) ) {
364 $this->loadItem( $code, $key );
365
366 return;
367 }
368
369 if ( !isset( $this->initialisedLangs[$code] ) ) {
370 $this->initLanguage( $code );
371 }
372
373 // Check to see if initLanguage() loaded it for us
374 if ( isset( $this->loadedItems[$code][$key] ) ||
375 isset( $this->loadedSubitems[$code][$key][$subkey] )
376 ) {
377 return;
378 }
379
380 if ( isset( $this->shallowFallbacks[$code] ) ) {
381 $this->loadSubitem( $this->shallowFallbacks[$code], $key, $subkey );
382
383 return;
384 }
385
386 $value = $this->store->get( $code, "$key:$subkey" );
387 $this->data[$code][$key][$subkey] = $value;
388 $this->loadedSubitems[$code][$key][$subkey] = true;
389 }
390
391 /**
392 * Returns true if the cache identified by $code is missing or expired.
393 *
394 * @param string $code
395 *
396 * @return bool
397 */
398 public function isExpired( $code ) {
399 if ( $this->forceRecache && !isset( $this->recachedLangs[$code] ) ) {
400 wfDebug( __METHOD__ . "($code): forced reload\n" );
401
402 return true;
403 }
404
405 $deps = $this->store->get( $code, 'deps' );
406 $keys = $this->store->get( $code, 'list' );
407 $preload = $this->store->get( $code, 'preload' );
408 // Different keys may expire separately for some stores
409 if ( $deps === null || $keys === null || $preload === null ) {
410 wfDebug( __METHOD__ . "($code): cache missing, need to make one\n" );
411
412 return true;
413 }
414
415 foreach ( $deps as $dep ) {
416 // Because we're unserializing stuff from cache, we
417 // could receive objects of classes that don't exist
418 // anymore (e.g. uninstalled extensions)
419 // When this happens, always expire the cache
420 if ( !$dep instanceof CacheDependency || $dep->isExpired() ) {
421 wfDebug( __METHOD__ . "($code): cache for $code expired due to " .
422 get_class( $dep ) . "\n" );
423
424 return true;
425 }
426 }
427
428 return false;
429 }
430
431 /**
432 * Initialise a language in this object. Rebuild the cache if necessary.
433 * @param string $code
434 * @throws MWException
435 */
436 protected function initLanguage( $code ) {
437 if ( isset( $this->initialisedLangs[$code] ) ) {
438 return;
439 }
440
441 $this->initialisedLangs[$code] = true;
442
443 # If the code is of the wrong form for a Messages*.php file, do a shallow fallback
444 if ( !Language::isValidBuiltInCode( $code ) ) {
445 $this->initShallowFallback( $code, 'en' );
446
447 return;
448 }
449
450 # Recache the data if necessary
451 if ( !$this->manualRecache && $this->isExpired( $code ) ) {
452 if ( Language::isSupportedLanguage( $code ) ) {
453 $this->recache( $code );
454 } elseif ( $code === 'en' ) {
455 throw new MWException( 'MessagesEn.php is missing.' );
456 } else {
457 $this->initShallowFallback( $code, 'en' );
458 }
459
460 return;
461 }
462
463 # Preload some stuff
464 $preload = $this->getItem( $code, 'preload' );
465 if ( $preload === null ) {
466 if ( $this->manualRecache ) {
467 // No Messages*.php file. Do shallow fallback to en.
468 if ( $code === 'en' ) {
469 throw new MWException( 'No localisation cache found for English. ' .
470 'Please run maintenance/rebuildLocalisationCache.php.' );
471 }
472 $this->initShallowFallback( $code, 'en' );
473
474 return;
475 } else {
476 throw new MWException( 'Invalid or missing localisation cache.' );
477 }
478 }
479 $this->data[$code] = $preload;
480 foreach ( $preload as $key => $item ) {
481 if ( in_array( $key, self::$splitKeys ) ) {
482 foreach ( $item as $subkey => $subitem ) {
483 $this->loadedSubitems[$code][$key][$subkey] = true;
484 }
485 } else {
486 $this->loadedItems[$code][$key] = true;
487 }
488 }
489 }
490
491 /**
492 * Create a fallback from one language to another, without creating a
493 * complete persistent cache.
494 * @param string $primaryCode
495 * @param string $fallbackCode
496 */
497 public function initShallowFallback( $primaryCode, $fallbackCode ) {
498 $this->data[$primaryCode] =& $this->data[$fallbackCode];
499 $this->loadedItems[$primaryCode] =& $this->loadedItems[$fallbackCode];
500 $this->loadedSubitems[$primaryCode] =& $this->loadedSubitems[$fallbackCode];
501 $this->shallowFallbacks[$primaryCode] = $fallbackCode;
502 }
503
504 /**
505 * Read a PHP file containing localisation data.
506 * @param string $_fileName
507 * @param string $_fileType
508 * @throws MWException
509 * @return array
510 */
511 protected function readPHPFile( $_fileName, $_fileType ) {
512 wfProfileIn( __METHOD__ );
513 // Disable APC caching
514 wfSuppressWarnings();
515 $_apcEnabled = ini_set( 'apc.cache_by_default', '0' );
516 wfRestoreWarnings();
517
518 include $_fileName;
519
520 wfSuppressWarnings();
521 ini_set( 'apc.cache_by_default', $_apcEnabled );
522 wfRestoreWarnings();
523
524 if ( $_fileType == 'core' || $_fileType == 'extension' ) {
525 $data = compact( self::$allKeys );
526 } elseif ( $_fileType == 'aliases' ) {
527 $data = compact( 'aliases' );
528 } else {
529 wfProfileOut( __METHOD__ );
530 throw new MWException( __METHOD__ . ": Invalid file type: $_fileType" );
531 }
532 wfProfileOut( __METHOD__ );
533
534 return $data;
535 }
536
537 /**
538 * Read a JSON file containing localisation messages.
539 * @param string $fileName Name of file to read
540 * @throws MWException If there is a syntax error in the JSON file
541 * @return array Array with a 'messages' key, or empty array if the file doesn't exist
542 */
543 public function readJSONFile( $fileName ) {
544 wfProfileIn( __METHOD__ );
545
546 if ( !is_readable( $fileName ) ) {
547 wfProfileOut( __METHOD__ );
548
549 return array();
550 }
551
552 $json = file_get_contents( $fileName );
553 if ( $json === false ) {
554 wfProfileOut( __METHOD__ );
555
556 return array();
557 }
558
559 $data = FormatJson::decode( $json, true );
560 if ( $data === null ) {
561 wfProfileOut( __METHOD__ );
562
563 throw new MWException( __METHOD__ . ": Invalid JSON file: $fileName" );
564 }
565
566 // Remove keys starting with '@', they're reserved for metadata and non-message data
567 foreach ( $data as $key => $unused ) {
568 if ( $key === '' || $key[0] === '@' ) {
569 unset( $data[$key] );
570 }
571 }
572
573 wfProfileOut( __METHOD__ );
574
575 // The JSON format only supports messages, none of the other variables, so wrap the data
576 return array( 'messages' => $data );
577 }
578
579 /**
580 * Get the compiled plural rules for a given language from the XML files.
581 * @since 1.20
582 * @param string $code
583 * @return array|null
584 */
585 public function getCompiledPluralRules( $code ) {
586 $rules = $this->getPluralRules( $code );
587 if ( $rules === null ) {
588 return null;
589 }
590 try {
591 $compiledRules = CLDRPluralRuleEvaluator::compile( $rules );
592 } catch ( CLDRPluralRuleError $e ) {
593 wfDebugLog( 'l10n', $e->getMessage() );
594
595 return array();
596 }
597
598 return $compiledRules;
599 }
600
601 /**
602 * Get the plural rules for a given language from the XML files.
603 * Cached.
604 * @since 1.20
605 * @param string $code
606 * @return array|null
607 */
608 public function getPluralRules( $code ) {
609 if ( $this->pluralRules === null ) {
610 $this->loadPluralFiles();
611 }
612 if ( !isset( $this->pluralRules[$code] ) ) {
613 return null;
614 } else {
615 return $this->pluralRules[$code];
616 }
617 }
618
619 /**
620 * Get the plural rule types for a given language from the XML files.
621 * Cached.
622 * @since 1.22
623 * @param string $code
624 * @return array|null
625 */
626 public function getPluralRuleTypes( $code ) {
627 if ( $this->pluralRuleTypes === null ) {
628 $this->loadPluralFiles();
629 }
630 if ( !isset( $this->pluralRuleTypes[$code] ) ) {
631 return null;
632 } else {
633 return $this->pluralRuleTypes[$code];
634 }
635 }
636
637 /**
638 * Load the plural XML files.
639 */
640 protected function loadPluralFiles() {
641 global $IP;
642 $cldrPlural = "$IP/languages/data/plurals.xml";
643 $mwPlural = "$IP/languages/data/plurals-mediawiki.xml";
644 // Load CLDR plural rules
645 $this->loadPluralFile( $cldrPlural );
646 if ( file_exists( $mwPlural ) ) {
647 // Override or extend
648 $this->loadPluralFile( $mwPlural );
649 }
650 }
651
652 /**
653 * Load a plural XML file with the given filename, compile the relevant
654 * rules, and save the compiled rules in a process-local cache.
655 *
656 * @param string $fileName
657 */
658 protected function loadPluralFile( $fileName ) {
659 // Use file_get_contents instead of DOMDocument::load (T58439)
660 $xml = file_get_contents( $fileName );
661 if ( !$xml ) {
662 throw new MWException( "Unable to read plurals file $fileName" );
663 }
664 $doc = new DOMDocument;
665 $doc->loadXML( $xml );
666 $rulesets = $doc->getElementsByTagName( "pluralRules" );
667 foreach ( $rulesets as $ruleset ) {
668 $codes = $ruleset->getAttribute( 'locales' );
669 $rules = array();
670 $ruleTypes = array();
671 $ruleElements = $ruleset->getElementsByTagName( "pluralRule" );
672 foreach ( $ruleElements as $elt ) {
673 $ruleType = $elt->getAttribute( 'count' );
674 if ( $ruleType === 'other' ) {
675 // Don't record "other" rules, which have an empty condition
676 continue;
677 }
678 $rules[] = $elt->nodeValue;
679 $ruleTypes[] = $ruleType;
680 }
681 foreach ( explode( ' ', $codes ) as $code ) {
682 $this->pluralRules[$code] = $rules;
683 $this->pluralRuleTypes[$code] = $ruleTypes;
684 }
685 }
686 }
687
688 /**
689 * Read the data from the source files for a given language, and register
690 * the relevant dependencies in the $deps array. If the localisation
691 * exists, the data array is returned, otherwise false is returned.
692 *
693 * @param string $code
694 * @param array $deps
695 * @return array
696 */
697 protected function readSourceFilesAndRegisterDeps( $code, &$deps ) {
698 global $IP;
699 wfProfileIn( __METHOD__ );
700
701 // This reads in the PHP i18n file with non-messages l10n data
702 $fileName = Language::getMessagesFileName( $code );
703 if ( !file_exists( $fileName ) ) {
704 $data = array();
705 } else {
706 $deps[] = new FileDependency( $fileName );
707 $data = $this->readPHPFile( $fileName, 'core' );
708 }
709
710 # Load CLDR plural rules for JavaScript
711 $data['pluralRules'] = $this->getPluralRules( $code );
712 # And for PHP
713 $data['compiledPluralRules'] = $this->getCompiledPluralRules( $code );
714 # Load plural rule types
715 $data['pluralRuleTypes'] = $this->getPluralRuleTypes( $code );
716
717 $deps['plurals'] = new FileDependency( "$IP/languages/data/plurals.xml" );
718 $deps['plurals-mw'] = new FileDependency( "$IP/languages/data/plurals-mediawiki.xml" );
719
720 wfProfileOut( __METHOD__ );
721
722 return $data;
723 }
724
725 /**
726 * Merge two localisation values, a primary and a fallback, overwriting the
727 * primary value in place.
728 * @param string $key
729 * @param mixed $value
730 * @param mixed $fallbackValue
731 */
732 protected function mergeItem( $key, &$value, $fallbackValue ) {
733 if ( !is_null( $value ) ) {
734 if ( !is_null( $fallbackValue ) ) {
735 if ( in_array( $key, self::$mergeableMapKeys ) ) {
736 $value = $value + $fallbackValue;
737 } elseif ( in_array( $key, self::$mergeableListKeys ) ) {
738 $value = array_unique( array_merge( $fallbackValue, $value ) );
739 } elseif ( in_array( $key, self::$mergeableAliasListKeys ) ) {
740 $value = array_merge_recursive( $value, $fallbackValue );
741 } elseif ( in_array( $key, self::$optionalMergeKeys ) ) {
742 if ( !empty( $value['inherit'] ) ) {
743 $value = array_merge( $fallbackValue, $value );
744 }
745
746 if ( isset( $value['inherit'] ) ) {
747 unset( $value['inherit'] );
748 }
749 } elseif ( in_array( $key, self::$magicWordKeys ) ) {
750 $this->mergeMagicWords( $value, $fallbackValue );
751 }
752 }
753 } else {
754 $value = $fallbackValue;
755 }
756 }
757
758 /**
759 * @param mixed $value
760 * @param mixed $fallbackValue
761 */
762 protected function mergeMagicWords( &$value, $fallbackValue ) {
763 foreach ( $fallbackValue as $magicName => $fallbackInfo ) {
764 if ( !isset( $value[$magicName] ) ) {
765 $value[$magicName] = $fallbackInfo;
766 } else {
767 $oldSynonyms = array_slice( $fallbackInfo, 1 );
768 $newSynonyms = array_slice( $value[$magicName], 1 );
769 $synonyms = array_values( array_unique( array_merge(
770 $newSynonyms, $oldSynonyms ) ) );
771 $value[$magicName] = array_merge( array( $fallbackInfo[0] ), $synonyms );
772 }
773 }
774 }
775
776 /**
777 * Given an array mapping language code to localisation value, such as is
778 * found in extension *.i18n.php files, iterate through a fallback sequence
779 * to merge the given data with an existing primary value.
780 *
781 * Returns true if any data from the extension array was used, false
782 * otherwise.
783 * @param array $codeSequence
784 * @param string $key
785 * @param mixed $value
786 * @param mixed $fallbackValue
787 * @return bool
788 */
789 protected function mergeExtensionItem( $codeSequence, $key, &$value, $fallbackValue ) {
790 $used = false;
791 foreach ( $codeSequence as $code ) {
792 if ( isset( $fallbackValue[$code] ) ) {
793 $this->mergeItem( $key, $value, $fallbackValue[$code] );
794 $used = true;
795 }
796 }
797
798 return $used;
799 }
800
801 /**
802 * Gets the combined list of messages dirs from
803 * core and extensions
804 *
805 * @since 1.25
806 * @return array
807 */
808 public function getMessagesDirs() {
809 global $wgMessagesDirs, $IP;
810 return array(
811 'core' => "$IP/languages/i18n",
812 'api' => "$IP/includes/api/i18n",
813 'oojs-ui' => "$IP/resources/lib/oojs-ui/i18n",
814 ) + $wgMessagesDirs;
815 }
816
817 /**
818 * Load localisation data for a given language for both core and extensions
819 * and save it to the persistent cache store and the process cache
820 * @param string $code
821 * @throws MWException
822 */
823 public function recache( $code ) {
824 global $wgExtensionMessagesFiles;
825 wfProfileIn( __METHOD__ );
826
827 if ( !$code ) {
828 wfProfileOut( __METHOD__ );
829 throw new MWException( "Invalid language code requested" );
830 }
831 $this->recachedLangs[$code] = true;
832
833 # Initial values
834 $initialData = array_combine(
835 self::$allKeys,
836 array_fill( 0, count( self::$allKeys ), null ) );
837 $coreData = $initialData;
838 $deps = array();
839
840 # Load the primary localisation from the source file
841 $data = $this->readSourceFilesAndRegisterDeps( $code, $deps );
842 if ( $data === false ) {
843 wfDebug( __METHOD__ . ": no localisation file for $code, using fallback to en\n" );
844 $coreData['fallback'] = 'en';
845 } else {
846 wfDebug( __METHOD__ . ": got localisation for $code from source\n" );
847
848 # Merge primary localisation
849 foreach ( $data as $key => $value ) {
850 $this->mergeItem( $key, $coreData[$key], $value );
851 }
852 }
853
854 # Fill in the fallback if it's not there already
855 if ( is_null( $coreData['fallback'] ) ) {
856 $coreData['fallback'] = $code === 'en' ? false : 'en';
857 }
858 if ( $coreData['fallback'] === false ) {
859 $coreData['fallbackSequence'] = array();
860 } else {
861 $coreData['fallbackSequence'] = array_map( 'trim', explode( ',', $coreData['fallback'] ) );
862 $len = count( $coreData['fallbackSequence'] );
863
864 # Ensure that the sequence ends at en
865 if ( $coreData['fallbackSequence'][$len - 1] !== 'en' ) {
866 $coreData['fallbackSequence'][] = 'en';
867 }
868 }
869
870 $codeSequence = array_merge( array( $code ), $coreData['fallbackSequence'] );
871 $messageDirs = $this->getMessagesDirs();
872
873 wfProfileIn( __METHOD__ . '-fallbacks' );
874
875 # Load non-JSON localisation data for extensions
876 $extensionData = array_combine(
877 $codeSequence,
878 array_fill( 0, count( $codeSequence ), $initialData ) );
879 foreach ( $wgExtensionMessagesFiles as $extension => $fileName ) {
880 if ( isset( $messageDirs[$extension] ) ) {
881 # This extension has JSON message data; skip the PHP shim
882 continue;
883 }
884
885 $data = $this->readPHPFile( $fileName, 'extension' );
886 $used = false;
887
888 foreach ( $data as $key => $item ) {
889 foreach ( $codeSequence as $csCode ) {
890 if ( isset( $item[$csCode] ) ) {
891 $this->mergeItem( $key, $extensionData[$csCode][$key], $item[$csCode] );
892 $used = true;
893 }
894 }
895 }
896
897 if ( $used ) {
898 $deps[] = new FileDependency( $fileName );
899 }
900 }
901
902 # Load the localisation data for each fallback, then merge it into the full array
903 $allData = $initialData;
904 foreach ( $codeSequence as $csCode ) {
905 $csData = $initialData;
906
907 # Load core messages and the extension localisations.
908 foreach ( $messageDirs as $dirs ) {
909 foreach ( (array)$dirs as $dir ) {
910 $fileName = "$dir/$csCode.json";
911 $data = $this->readJSONFile( $fileName );
912
913 foreach ( $data as $key => $item ) {
914 $this->mergeItem( $key, $csData[$key], $item );
915 }
916
917 $deps[] = new FileDependency( $fileName );
918 }
919 }
920
921 # Merge non-JSON extension data
922 if ( isset( $extensionData[$csCode] ) ) {
923 foreach ( $extensionData[$csCode] as $key => $item ) {
924 $this->mergeItem( $key, $csData[$key], $item );
925 }
926 }
927
928 if ( $csCode === $code ) {
929 # Merge core data into extension data
930 foreach ( $coreData as $key => $item ) {
931 $this->mergeItem( $key, $csData[$key], $item );
932 }
933 } else {
934 # Load the secondary localisation from the source file to
935 # avoid infinite cycles on cyclic fallbacks
936 $fbData = $this->readSourceFilesAndRegisterDeps( $csCode, $deps );
937 if ( $fbData !== false ) {
938 # Only merge the keys that make sense to merge
939 foreach ( self::$allKeys as $key ) {
940 if ( !isset( $fbData[$key] ) ) {
941 continue;
942 }
943
944 if ( is_null( $coreData[$key] ) || $this->isMergeableKey( $key ) ) {
945 $this->mergeItem( $key, $csData[$key], $fbData[$key] );
946 }
947 }
948 }
949 }
950
951 # Allow extensions an opportunity to adjust the data for this
952 # fallback
953 Hooks::run( 'LocalisationCacheRecacheFallback', array( $this, $csCode, &$csData ) );
954
955 # Merge the data for this fallback into the final array
956 if ( $csCode === $code ) {
957 $allData = $csData;
958 } else {
959 foreach ( self::$allKeys as $key ) {
960 if ( !isset( $csData[$key] ) ) {
961 continue;
962 }
963
964 if ( is_null( $allData[$key] ) || $this->isMergeableKey( $key ) ) {
965 $this->mergeItem( $key, $allData[$key], $csData[$key] );
966 }
967 }
968 }
969 }
970
971 wfProfileOut( __METHOD__ . '-fallbacks' );
972
973 # Add cache dependencies for any referenced globals
974 $deps['wgExtensionMessagesFiles'] = new GlobalDependency( 'wgExtensionMessagesFiles' );
975 // $wgMessagesDirs is used in LocalisationCache::getMessagesDirs()
976 $deps['wgMessagesDirs'] = new GlobalDependency( 'wgMessagesDirs' );
977 $deps['version'] = new ConstantDependency( 'LocalisationCache::VERSION' );
978
979 # Add dependencies to the cache entry
980 $allData['deps'] = $deps;
981
982 # Replace spaces with underscores in namespace names
983 $allData['namespaceNames'] = str_replace( ' ', '_', $allData['namespaceNames'] );
984
985 # And do the same for special page aliases. $page is an array.
986 foreach ( $allData['specialPageAliases'] as &$page ) {
987 $page = str_replace( ' ', '_', $page );
988 }
989 # Decouple the reference to prevent accidental damage
990 unset( $page );
991
992 # If there were no plural rules, return an empty array
993 if ( $allData['pluralRules'] === null ) {
994 $allData['pluralRules'] = array();
995 }
996 if ( $allData['compiledPluralRules'] === null ) {
997 $allData['compiledPluralRules'] = array();
998 }
999 # If there were no plural rule types, return an empty array
1000 if ( $allData['pluralRuleTypes'] === null ) {
1001 $allData['pluralRuleTypes'] = array();
1002 }
1003
1004 # Set the list keys
1005 $allData['list'] = array();
1006 foreach ( self::$splitKeys as $key ) {
1007 $allData['list'][$key] = array_keys( $allData[$key] );
1008 }
1009 # Run hooks
1010 $purgeBlobs = true;
1011 Hooks::run( 'LocalisationCacheRecache', array( $this, $code, &$allData, &$purgeBlobs ) );
1012
1013 if ( is_null( $allData['namespaceNames'] ) ) {
1014 wfProfileOut( __METHOD__ );
1015 throw new MWException( __METHOD__ . ': Localisation data failed sanity check! ' .
1016 'Check that your languages/messages/MessagesEn.php file is intact.' );
1017 }
1018
1019 # Set the preload key
1020 $allData['preload'] = $this->buildPreload( $allData );
1021
1022 # Save to the process cache and register the items loaded
1023 $this->data[$code] = $allData;
1024 foreach ( $allData as $key => $item ) {
1025 $this->loadedItems[$code][$key] = true;
1026 }
1027
1028 # Save to the persistent cache
1029 wfProfileIn( __METHOD__ . '-write' );
1030 $this->store->startWrite( $code );
1031 foreach ( $allData as $key => $value ) {
1032 if ( in_array( $key, self::$splitKeys ) ) {
1033 foreach ( $value as $subkey => $subvalue ) {
1034 $this->store->set( "$key:$subkey", $subvalue );
1035 }
1036 } else {
1037 $this->store->set( $key, $value );
1038 }
1039 }
1040 $this->store->finishWrite();
1041 wfProfileOut( __METHOD__ . '-write' );
1042
1043 # Clear out the MessageBlobStore
1044 # HACK: If using a null (i.e. disabled) storage backend, we
1045 # can't write to the MessageBlobStore either
1046 if ( $purgeBlobs && !$this->store instanceof LCStoreNull ) {
1047 MessageBlobStore::getInstance()->clear();
1048 }
1049
1050 wfProfileOut( __METHOD__ );
1051 }
1052
1053 /**
1054 * Build the preload item from the given pre-cache data.
1055 *
1056 * The preload item will be loaded automatically, improving performance
1057 * for the commonly-requested items it contains.
1058 * @param array $data
1059 * @return array
1060 */
1061 protected function buildPreload( $data ) {
1062 $preload = array( 'messages' => array() );
1063 foreach ( self::$preloadedKeys as $key ) {
1064 $preload[$key] = $data[$key];
1065 }
1066
1067 foreach ( $data['preloadedMessages'] as $subkey ) {
1068 if ( isset( $data['messages'][$subkey] ) ) {
1069 $subitem = $data['messages'][$subkey];
1070 } else {
1071 $subitem = null;
1072 }
1073 $preload['messages'][$subkey] = $subitem;
1074 }
1075
1076 return $preload;
1077 }
1078
1079 /**
1080 * Unload the data for a given language from the object cache.
1081 * Reduces memory usage.
1082 * @param string $code
1083 */
1084 public function unload( $code ) {
1085 unset( $this->data[$code] );
1086 unset( $this->loadedItems[$code] );
1087 unset( $this->loadedSubitems[$code] );
1088 unset( $this->initialisedLangs[$code] );
1089 unset( $this->shallowFallbacks[$code] );
1090
1091 foreach ( $this->shallowFallbacks as $shallowCode => $fbCode ) {
1092 if ( $fbCode === $code ) {
1093 $this->unload( $shallowCode );
1094 }
1095 }
1096 }
1097
1098 /**
1099 * Unload all data
1100 */
1101 public function unloadAll() {
1102 foreach ( $this->initialisedLangs as $lang => $unused ) {
1103 $this->unload( $lang );
1104 }
1105 }
1106
1107 /**
1108 * Disable the storage backend
1109 */
1110 public function disableBackend() {
1111 $this->store = new LCStoreNull;
1112 $this->manualRecache = false;
1113 }
1114 }
1115
1116 /**
1117 * Interface for the persistence layer of LocalisationCache.
1118 *
1119 * The persistence layer is two-level hierarchical cache. The first level
1120 * is the language, the second level is the item or subitem.
1121 *
1122 * Since the data for a whole language is rebuilt in one operation, it needs
1123 * to have a fast and atomic method for deleting or replacing all of the
1124 * current data for a given language. The interface reflects this bulk update
1125 * operation. Callers writing to the cache must first call startWrite(), then
1126 * will call set() a couple of thousand times, then will call finishWrite()
1127 * to commit the operation. When finishWrite() is called, the cache is
1128 * expected to delete all data previously stored for that language.
1129 *
1130 * The values stored are PHP variables suitable for serialize(). Implementations
1131 * of LCStore are responsible for serializing and unserializing.
1132 */
1133 interface LCStore {
1134 /**
1135 * Get a value.
1136 * @param string $code Language code
1137 * @param string $key Cache key
1138 */
1139 function get( $code, $key );
1140
1141 /**
1142 * Start a write transaction.
1143 * @param string $code Language code
1144 */
1145 function startWrite( $code );
1146
1147 /**
1148 * Finish a write transaction.
1149 */
1150 function finishWrite();
1151
1152 /**
1153 * Set a key to a given value. startWrite() must be called before this
1154 * is called, and finishWrite() must be called afterwards.
1155 * @param string $key
1156 * @param mixed $value
1157 */
1158 function set( $key, $value );
1159 }
1160
1161 /**
1162 * LCStore implementation which uses the standard DB functions to store data.
1163 * This will work on any MediaWiki installation.
1164 */
1165 class LCStoreDB implements LCStore {
1166 private $currentLang;
1167 private $writesDone = false;
1168
1169 /** @var DatabaseBase */
1170 private $dbw;
1171 /** @var array */
1172 private $batch = array();
1173
1174 private $readOnly = false;
1175
1176 public function get( $code, $key ) {
1177 if ( $this->writesDone ) {
1178 $db = wfGetDB( DB_MASTER );
1179 } else {
1180 $db = wfGetDB( DB_SLAVE );
1181 }
1182 $row = $db->selectRow( 'l10n_cache', array( 'lc_value' ),
1183 array( 'lc_lang' => $code, 'lc_key' => $key ), __METHOD__ );
1184 if ( $row ) {
1185 return unserialize( $db->decodeBlob( $row->lc_value ) );
1186 } else {
1187 return null;
1188 }
1189 }
1190
1191 public function startWrite( $code ) {
1192 if ( $this->readOnly ) {
1193 return;
1194 } elseif ( !$code ) {
1195 throw new MWException( __METHOD__ . ": Invalid language \"$code\"" );
1196 }
1197
1198 $this->dbw = wfGetDB( DB_MASTER );
1199
1200 $this->currentLang = $code;
1201 $this->batch = array();
1202 }
1203
1204 public function finishWrite() {
1205 if ( $this->readOnly ) {
1206 return;
1207 } elseif ( is_null( $this->currentLang ) ) {
1208 throw new MWException( __CLASS__ . ': must call startWrite() before finishWrite()' );
1209 }
1210
1211 $this->dbw->begin( __METHOD__ );
1212 try {
1213 $this->dbw->delete( 'l10n_cache',
1214 array( 'lc_lang' => $this->currentLang ), __METHOD__ );
1215 foreach ( array_chunk( $this->batch, 500 ) as $rows ) {
1216 $this->dbw->insert( 'l10n_cache', $rows, __METHOD__ );
1217 }
1218 $this->writesDone = true;
1219 } catch ( DBQueryError $e ) {
1220 if ( $this->dbw->wasReadOnlyError() ) {
1221 $this->readOnly = true; // just avoid site down time
1222 } else {
1223 throw $e;
1224 }
1225 }
1226 $this->dbw->commit( __METHOD__ );
1227
1228 $this->currentLang = null;
1229 $this->batch = array();
1230 }
1231
1232 public function set( $key, $value ) {
1233 if ( $this->readOnly ) {
1234 return;
1235 } elseif ( is_null( $this->currentLang ) ) {
1236 throw new MWException( __CLASS__ . ': must call startWrite() before set()' );
1237 }
1238
1239 $this->batch[] = array(
1240 'lc_lang' => $this->currentLang,
1241 'lc_key' => $key,
1242 'lc_value' => $this->dbw->encodeBlob( serialize( $value ) ) );
1243 }
1244 }
1245
1246 /**
1247 * LCStore implementation which stores data as a collection of CDB files in the
1248 * directory given by $wgCacheDirectory. If $wgCacheDirectory is not set, this
1249 * will throw an exception.
1250 *
1251 * Profiling indicates that on Linux, this implementation outperforms MySQL if
1252 * the directory is on a local filesystem and there is ample kernel cache
1253 * space. The performance advantage is greater when the DBA extension is
1254 * available than it is with the PHP port.
1255 *
1256 * See Cdb.php and http://cr.yp.to/cdb.html
1257 */
1258 class LCStoreCDB implements LCStore {
1259 /** @var CdbReader[] */
1260 private $readers;
1261
1262 /** @var CdbWriter */
1263 private $writer;
1264
1265 /** @var string Current language code */
1266 private $currentLang;
1267
1268 /** @var bool|string Cache directory. False if not set */
1269 private $directory;
1270
1271 function __construct( $conf = array() ) {
1272 global $wgCacheDirectory;
1273
1274 if ( isset( $conf['directory'] ) ) {
1275 $this->directory = $conf['directory'];
1276 } else {
1277 $this->directory = $wgCacheDirectory;
1278 }
1279 }
1280
1281 public function get( $code, $key ) {
1282 if ( !isset( $this->readers[$code] ) ) {
1283 $fileName = $this->getFileName( $code );
1284
1285 $this->readers[$code] = false;
1286 if ( file_exists( $fileName ) ) {
1287 try {
1288 $this->readers[$code] = CdbReader::open( $fileName );
1289 } catch ( CdbException $e ) {
1290 wfDebug( __METHOD__ . ": unable to open cdb file for reading\n" );
1291 }
1292 }
1293 }
1294
1295 if ( !$this->readers[$code] ) {
1296 return null;
1297 } else {
1298 $value = false;
1299 try {
1300 $value = $this->readers[$code]->get( $key );
1301 } catch ( CdbException $e ) {
1302 wfDebug( __METHOD__ . ": CdbException caught, error message was "
1303 . $e->getMessage() . "\n" );
1304 }
1305 if ( $value === false ) {
1306 return null;
1307 }
1308
1309 return unserialize( $value );
1310 }
1311 }
1312
1313 public function startWrite( $code ) {
1314 if ( !file_exists( $this->directory ) ) {
1315 if ( !wfMkdirParents( $this->directory, null, __METHOD__ ) ) {
1316 throw new MWException( "Unable to create the localisation store " .
1317 "directory \"{$this->directory}\"" );
1318 }
1319 }
1320
1321 // Close reader to stop permission errors on write
1322 if ( !empty( $this->readers[$code] ) ) {
1323 $this->readers[$code]->close();
1324 }
1325
1326 try {
1327 $this->writer = CdbWriter::open( $this->getFileName( $code ) );
1328 } catch ( CdbException $e ) {
1329 throw new MWException( $e->getMessage() );
1330 }
1331 $this->currentLang = $code;
1332 }
1333
1334 public function finishWrite() {
1335 // Close the writer
1336 try {
1337 $this->writer->close();
1338 } catch ( CdbException $e ) {
1339 throw new MWException( $e->getMessage() );
1340 }
1341 $this->writer = null;
1342 unset( $this->readers[$this->currentLang] );
1343 $this->currentLang = null;
1344 }
1345
1346 public function set( $key, $value ) {
1347 if ( is_null( $this->writer ) ) {
1348 throw new MWException( __CLASS__ . ': must call startWrite() before calling set()' );
1349 }
1350 try {
1351 $this->writer->set( $key, serialize( $value ) );
1352 } catch ( CdbException $e ) {
1353 throw new MWException( $e->getMessage() );
1354 }
1355 }
1356
1357 protected function getFileName( $code ) {
1358 if ( strval( $code ) === '' || strpos( $code, '/' ) !== false ) {
1359 throw new MWException( __METHOD__ . ": Invalid language \"$code\"" );
1360 }
1361
1362 return "{$this->directory}/l10n_cache-$code.cdb";
1363 }
1364 }
1365
1366 /**
1367 * Null store backend, used to avoid DB errors during install
1368 */
1369 class LCStoreNull implements LCStore {
1370 public function get( $code, $key ) {
1371 return null;
1372 }
1373
1374 public function startWrite( $code ) {
1375 }
1376
1377 public function finishWrite() {
1378 }
1379
1380 public function set( $key, $value ) {
1381 }
1382 }
1383
1384 /**
1385 * A localisation cache optimised for loading large amounts of data for many
1386 * languages. Used by rebuildLocalisationCache.php.
1387 */
1388 class LocalisationCacheBulkLoad extends LocalisationCache {
1389 /**
1390 * A cache of the contents of data files.
1391 * Core files are serialized to avoid using ~1GB of RAM during a recache.
1392 */
1393 private $fileCache = array();
1394
1395 /**
1396 * Most recently used languages. Uses the linked-list aspect of PHP hashtables
1397 * to keep the most recently used language codes at the end of the array, and
1398 * the language codes that are ready to be deleted at the beginning.
1399 */
1400 private $mruLangs = array();
1401
1402 /**
1403 * Maximum number of languages that may be loaded into $this->data
1404 */
1405 private $maxLoadedLangs = 10;
1406
1407 /**
1408 * @param string $fileName
1409 * @param string $fileType
1410 * @return array|mixed
1411 */
1412 protected function readPHPFile( $fileName, $fileType ) {
1413 $serialize = $fileType === 'core';
1414 if ( !isset( $this->fileCache[$fileName][$fileType] ) ) {
1415 $data = parent::readPHPFile( $fileName, $fileType );
1416
1417 if ( $serialize ) {
1418 $encData = serialize( $data );
1419 } else {
1420 $encData = $data;
1421 }
1422
1423 $this->fileCache[$fileName][$fileType] = $encData;
1424
1425 return $data;
1426 } elseif ( $serialize ) {
1427 return unserialize( $this->fileCache[$fileName][$fileType] );
1428 } else {
1429 return $this->fileCache[$fileName][$fileType];
1430 }
1431 }
1432
1433 /**
1434 * @param string $code
1435 * @param string $key
1436 * @return mixed
1437 */
1438 public function getItem( $code, $key ) {
1439 unset( $this->mruLangs[$code] );
1440 $this->mruLangs[$code] = true;
1441
1442 return parent::getItem( $code, $key );
1443 }
1444
1445 /**
1446 * @param string $code
1447 * @param string $key
1448 * @param string $subkey
1449 * @return mixed
1450 */
1451 public function getSubitem( $code, $key, $subkey ) {
1452 unset( $this->mruLangs[$code] );
1453 $this->mruLangs[$code] = true;
1454
1455 return parent::getSubitem( $code, $key, $subkey );
1456 }
1457
1458 /**
1459 * @param string $code
1460 */
1461 public function recache( $code ) {
1462 parent::recache( $code );
1463 unset( $this->mruLangs[$code] );
1464 $this->mruLangs[$code] = true;
1465 $this->trimCache();
1466 }
1467
1468 /**
1469 * @param string $code
1470 */
1471 public function unload( $code ) {
1472 unset( $this->mruLangs[$code] );
1473 parent::unload( $code );
1474 }
1475
1476 /**
1477 * Unload cached languages until there are less than $this->maxLoadedLangs
1478 */
1479 protected function trimCache() {
1480 while ( count( $this->data ) > $this->maxLoadedLangs && count( $this->mruLangs ) ) {
1481 reset( $this->mruLangs );
1482 $code = key( $this->mruLangs );
1483 wfDebug( __METHOD__ . ": unloading $code\n" );
1484 $this->unload( $code );
1485 }
1486 }
1487 }