DatabaseMssql: Don't duplicate body of makeList()
[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 * @throws MWException
658 */
659 protected function loadPluralFile( $fileName ) {
660 // Use file_get_contents instead of DOMDocument::load (T58439)
661 $xml = file_get_contents( $fileName );
662 if ( !$xml ) {
663 throw new MWException( "Unable to read plurals file $fileName" );
664 }
665 $doc = new DOMDocument;
666 $doc->loadXML( $xml );
667 $rulesets = $doc->getElementsByTagName( "pluralRules" );
668 foreach ( $rulesets as $ruleset ) {
669 $codes = $ruleset->getAttribute( 'locales' );
670 $rules = array();
671 $ruleTypes = array();
672 $ruleElements = $ruleset->getElementsByTagName( "pluralRule" );
673 foreach ( $ruleElements as $elt ) {
674 $ruleType = $elt->getAttribute( 'count' );
675 if ( $ruleType === 'other' ) {
676 // Don't record "other" rules, which have an empty condition
677 continue;
678 }
679 $rules[] = $elt->nodeValue;
680 $ruleTypes[] = $ruleType;
681 }
682 foreach ( explode( ' ', $codes ) as $code ) {
683 $this->pluralRules[$code] = $rules;
684 $this->pluralRuleTypes[$code] = $ruleTypes;
685 }
686 }
687 }
688
689 /**
690 * Read the data from the source files for a given language, and register
691 * the relevant dependencies in the $deps array. If the localisation
692 * exists, the data array is returned, otherwise false is returned.
693 *
694 * @param string $code
695 * @param array $deps
696 * @return array
697 */
698 protected function readSourceFilesAndRegisterDeps( $code, &$deps ) {
699 global $IP;
700 wfProfileIn( __METHOD__ );
701
702 // This reads in the PHP i18n file with non-messages l10n data
703 $fileName = Language::getMessagesFileName( $code );
704 if ( !file_exists( $fileName ) ) {
705 $data = array();
706 } else {
707 $deps[] = new FileDependency( $fileName );
708 $data = $this->readPHPFile( $fileName, 'core' );
709 }
710
711 # Load CLDR plural rules for JavaScript
712 $data['pluralRules'] = $this->getPluralRules( $code );
713 # And for PHP
714 $data['compiledPluralRules'] = $this->getCompiledPluralRules( $code );
715 # Load plural rule types
716 $data['pluralRuleTypes'] = $this->getPluralRuleTypes( $code );
717
718 $deps['plurals'] = new FileDependency( "$IP/languages/data/plurals.xml" );
719 $deps['plurals-mw'] = new FileDependency( "$IP/languages/data/plurals-mediawiki.xml" );
720
721 wfProfileOut( __METHOD__ );
722
723 return $data;
724 }
725
726 /**
727 * Merge two localisation values, a primary and a fallback, overwriting the
728 * primary value in place.
729 * @param string $key
730 * @param mixed $value
731 * @param mixed $fallbackValue
732 */
733 protected function mergeItem( $key, &$value, $fallbackValue ) {
734 if ( !is_null( $value ) ) {
735 if ( !is_null( $fallbackValue ) ) {
736 if ( in_array( $key, self::$mergeableMapKeys ) ) {
737 $value = $value + $fallbackValue;
738 } elseif ( in_array( $key, self::$mergeableListKeys ) ) {
739 $value = array_unique( array_merge( $fallbackValue, $value ) );
740 } elseif ( in_array( $key, self::$mergeableAliasListKeys ) ) {
741 $value = array_merge_recursive( $value, $fallbackValue );
742 } elseif ( in_array( $key, self::$optionalMergeKeys ) ) {
743 if ( !empty( $value['inherit'] ) ) {
744 $value = array_merge( $fallbackValue, $value );
745 }
746
747 if ( isset( $value['inherit'] ) ) {
748 unset( $value['inherit'] );
749 }
750 } elseif ( in_array( $key, self::$magicWordKeys ) ) {
751 $this->mergeMagicWords( $value, $fallbackValue );
752 }
753 }
754 } else {
755 $value = $fallbackValue;
756 }
757 }
758
759 /**
760 * @param mixed $value
761 * @param mixed $fallbackValue
762 */
763 protected function mergeMagicWords( &$value, $fallbackValue ) {
764 foreach ( $fallbackValue as $magicName => $fallbackInfo ) {
765 if ( !isset( $value[$magicName] ) ) {
766 $value[$magicName] = $fallbackInfo;
767 } else {
768 $oldSynonyms = array_slice( $fallbackInfo, 1 );
769 $newSynonyms = array_slice( $value[$magicName], 1 );
770 $synonyms = array_values( array_unique( array_merge(
771 $newSynonyms, $oldSynonyms ) ) );
772 $value[$magicName] = array_merge( array( $fallbackInfo[0] ), $synonyms );
773 }
774 }
775 }
776
777 /**
778 * Given an array mapping language code to localisation value, such as is
779 * found in extension *.i18n.php files, iterate through a fallback sequence
780 * to merge the given data with an existing primary value.
781 *
782 * Returns true if any data from the extension array was used, false
783 * otherwise.
784 * @param array $codeSequence
785 * @param string $key
786 * @param mixed $value
787 * @param mixed $fallbackValue
788 * @return bool
789 */
790 protected function mergeExtensionItem( $codeSequence, $key, &$value, $fallbackValue ) {
791 $used = false;
792 foreach ( $codeSequence as $code ) {
793 if ( isset( $fallbackValue[$code] ) ) {
794 $this->mergeItem( $key, $value, $fallbackValue[$code] );
795 $used = true;
796 }
797 }
798
799 return $used;
800 }
801
802 /**
803 * Gets the combined list of messages dirs from
804 * core and extensions
805 *
806 * @since 1.25
807 * @return array
808 */
809 public function getMessagesDirs() {
810 global $wgMessagesDirs, $IP;
811 return array(
812 'core' => "$IP/languages/i18n",
813 'api' => "$IP/includes/api/i18n",
814 'oojs-ui' => "$IP/resources/lib/oojs-ui/i18n",
815 ) + $wgMessagesDirs;
816 }
817
818 /**
819 * Load localisation data for a given language for both core and extensions
820 * and save it to the persistent cache store and the process cache
821 * @param string $code
822 * @throws MWException
823 */
824 public function recache( $code ) {
825 global $wgExtensionMessagesFiles;
826 wfProfileIn( __METHOD__ );
827
828 if ( !$code ) {
829 wfProfileOut( __METHOD__ );
830 throw new MWException( "Invalid language code requested" );
831 }
832 $this->recachedLangs[$code] = true;
833
834 # Initial values
835 $initialData = array_combine(
836 self::$allKeys,
837 array_fill( 0, count( self::$allKeys ), null ) );
838 $coreData = $initialData;
839 $deps = array();
840
841 # Load the primary localisation from the source file
842 $data = $this->readSourceFilesAndRegisterDeps( $code, $deps );
843 if ( $data === false ) {
844 wfDebug( __METHOD__ . ": no localisation file for $code, using fallback to en\n" );
845 $coreData['fallback'] = 'en';
846 } else {
847 wfDebug( __METHOD__ . ": got localisation for $code from source\n" );
848
849 # Merge primary localisation
850 foreach ( $data as $key => $value ) {
851 $this->mergeItem( $key, $coreData[$key], $value );
852 }
853 }
854
855 # Fill in the fallback if it's not there already
856 if ( is_null( $coreData['fallback'] ) ) {
857 $coreData['fallback'] = $code === 'en' ? false : 'en';
858 }
859 if ( $coreData['fallback'] === false ) {
860 $coreData['fallbackSequence'] = array();
861 } else {
862 $coreData['fallbackSequence'] = array_map( 'trim', explode( ',', $coreData['fallback'] ) );
863 $len = count( $coreData['fallbackSequence'] );
864
865 # Ensure that the sequence ends at en
866 if ( $coreData['fallbackSequence'][$len - 1] !== 'en' ) {
867 $coreData['fallbackSequence'][] = 'en';
868 }
869 }
870
871 $codeSequence = array_merge( array( $code ), $coreData['fallbackSequence'] );
872 $messageDirs = $this->getMessagesDirs();
873
874 wfProfileIn( __METHOD__ . '-fallbacks' );
875
876 # Load non-JSON localisation data for extensions
877 $extensionData = array_combine(
878 $codeSequence,
879 array_fill( 0, count( $codeSequence ), $initialData ) );
880 foreach ( $wgExtensionMessagesFiles as $extension => $fileName ) {
881 if ( isset( $messageDirs[$extension] ) ) {
882 # This extension has JSON message data; skip the PHP shim
883 continue;
884 }
885
886 $data = $this->readPHPFile( $fileName, 'extension' );
887 $used = false;
888
889 foreach ( $data as $key => $item ) {
890 foreach ( $codeSequence as $csCode ) {
891 if ( isset( $item[$csCode] ) ) {
892 $this->mergeItem( $key, $extensionData[$csCode][$key], $item[$csCode] );
893 $used = true;
894 }
895 }
896 }
897
898 if ( $used ) {
899 $deps[] = new FileDependency( $fileName );
900 }
901 }
902
903 # Load the localisation data for each fallback, then merge it into the full array
904 $allData = $initialData;
905 foreach ( $codeSequence as $csCode ) {
906 $csData = $initialData;
907
908 # Load core messages and the extension localisations.
909 foreach ( $messageDirs as $dirs ) {
910 foreach ( (array)$dirs as $dir ) {
911 $fileName = "$dir/$csCode.json";
912 $data = $this->readJSONFile( $fileName );
913
914 foreach ( $data as $key => $item ) {
915 $this->mergeItem( $key, $csData[$key], $item );
916 }
917
918 $deps[] = new FileDependency( $fileName );
919 }
920 }
921
922 # Merge non-JSON extension data
923 if ( isset( $extensionData[$csCode] ) ) {
924 foreach ( $extensionData[$csCode] as $key => $item ) {
925 $this->mergeItem( $key, $csData[$key], $item );
926 }
927 }
928
929 if ( $csCode === $code ) {
930 # Merge core data into extension data
931 foreach ( $coreData as $key => $item ) {
932 $this->mergeItem( $key, $csData[$key], $item );
933 }
934 } else {
935 # Load the secondary localisation from the source file to
936 # avoid infinite cycles on cyclic fallbacks
937 $fbData = $this->readSourceFilesAndRegisterDeps( $csCode, $deps );
938 if ( $fbData !== false ) {
939 # Only merge the keys that make sense to merge
940 foreach ( self::$allKeys as $key ) {
941 if ( !isset( $fbData[$key] ) ) {
942 continue;
943 }
944
945 if ( is_null( $coreData[$key] ) || $this->isMergeableKey( $key ) ) {
946 $this->mergeItem( $key, $csData[$key], $fbData[$key] );
947 }
948 }
949 }
950 }
951
952 # Allow extensions an opportunity to adjust the data for this
953 # fallback
954 Hooks::run( 'LocalisationCacheRecacheFallback', array( $this, $csCode, &$csData ) );
955
956 # Merge the data for this fallback into the final array
957 if ( $csCode === $code ) {
958 $allData = $csData;
959 } else {
960 foreach ( self::$allKeys as $key ) {
961 if ( !isset( $csData[$key] ) ) {
962 continue;
963 }
964
965 if ( is_null( $allData[$key] ) || $this->isMergeableKey( $key ) ) {
966 $this->mergeItem( $key, $allData[$key], $csData[$key] );
967 }
968 }
969 }
970 }
971
972 wfProfileOut( __METHOD__ . '-fallbacks' );
973
974 # Add cache dependencies for any referenced globals
975 $deps['wgExtensionMessagesFiles'] = new GlobalDependency( 'wgExtensionMessagesFiles' );
976 // $wgMessagesDirs is used in LocalisationCache::getMessagesDirs()
977 $deps['wgMessagesDirs'] = new GlobalDependency( 'wgMessagesDirs' );
978 $deps['version'] = new ConstantDependency( 'LocalisationCache::VERSION' );
979
980 # Add dependencies to the cache entry
981 $allData['deps'] = $deps;
982
983 # Replace spaces with underscores in namespace names
984 $allData['namespaceNames'] = str_replace( ' ', '_', $allData['namespaceNames'] );
985
986 # And do the same for special page aliases. $page is an array.
987 foreach ( $allData['specialPageAliases'] as &$page ) {
988 $page = str_replace( ' ', '_', $page );
989 }
990 # Decouple the reference to prevent accidental damage
991 unset( $page );
992
993 # If there were no plural rules, return an empty array
994 if ( $allData['pluralRules'] === null ) {
995 $allData['pluralRules'] = array();
996 }
997 if ( $allData['compiledPluralRules'] === null ) {
998 $allData['compiledPluralRules'] = array();
999 }
1000 # If there were no plural rule types, return an empty array
1001 if ( $allData['pluralRuleTypes'] === null ) {
1002 $allData['pluralRuleTypes'] = array();
1003 }
1004
1005 # Set the list keys
1006 $allData['list'] = array();
1007 foreach ( self::$splitKeys as $key ) {
1008 $allData['list'][$key] = array_keys( $allData[$key] );
1009 }
1010 # Run hooks
1011 $purgeBlobs = true;
1012 Hooks::run( 'LocalisationCacheRecache', array( $this, $code, &$allData, &$purgeBlobs ) );
1013
1014 if ( is_null( $allData['namespaceNames'] ) ) {
1015 wfProfileOut( __METHOD__ );
1016 throw new MWException( __METHOD__ . ': Localisation data failed sanity check! ' .
1017 'Check that your languages/messages/MessagesEn.php file is intact.' );
1018 }
1019
1020 # Set the preload key
1021 $allData['preload'] = $this->buildPreload( $allData );
1022
1023 # Save to the process cache and register the items loaded
1024 $this->data[$code] = $allData;
1025 foreach ( $allData as $key => $item ) {
1026 $this->loadedItems[$code][$key] = true;
1027 }
1028
1029 # Save to the persistent cache
1030 wfProfileIn( __METHOD__ . '-write' );
1031 $this->store->startWrite( $code );
1032 foreach ( $allData as $key => $value ) {
1033 if ( in_array( $key, self::$splitKeys ) ) {
1034 foreach ( $value as $subkey => $subvalue ) {
1035 $this->store->set( "$key:$subkey", $subvalue );
1036 }
1037 } else {
1038 $this->store->set( $key, $value );
1039 }
1040 }
1041 $this->store->finishWrite();
1042 wfProfileOut( __METHOD__ . '-write' );
1043
1044 # Clear out the MessageBlobStore
1045 # HACK: If using a null (i.e. disabled) storage backend, we
1046 # can't write to the MessageBlobStore either
1047 if ( $purgeBlobs && !$this->store instanceof LCStoreNull ) {
1048 MessageBlobStore::getInstance()->clear();
1049 }
1050
1051 wfProfileOut( __METHOD__ );
1052 }
1053
1054 /**
1055 * Build the preload item from the given pre-cache data.
1056 *
1057 * The preload item will be loaded automatically, improving performance
1058 * for the commonly-requested items it contains.
1059 * @param array $data
1060 * @return array
1061 */
1062 protected function buildPreload( $data ) {
1063 $preload = array( 'messages' => array() );
1064 foreach ( self::$preloadedKeys as $key ) {
1065 $preload[$key] = $data[$key];
1066 }
1067
1068 foreach ( $data['preloadedMessages'] as $subkey ) {
1069 if ( isset( $data['messages'][$subkey] ) ) {
1070 $subitem = $data['messages'][$subkey];
1071 } else {
1072 $subitem = null;
1073 }
1074 $preload['messages'][$subkey] = $subitem;
1075 }
1076
1077 return $preload;
1078 }
1079
1080 /**
1081 * Unload the data for a given language from the object cache.
1082 * Reduces memory usage.
1083 * @param string $code
1084 */
1085 public function unload( $code ) {
1086 unset( $this->data[$code] );
1087 unset( $this->loadedItems[$code] );
1088 unset( $this->loadedSubitems[$code] );
1089 unset( $this->initialisedLangs[$code] );
1090 unset( $this->shallowFallbacks[$code] );
1091
1092 foreach ( $this->shallowFallbacks as $shallowCode => $fbCode ) {
1093 if ( $fbCode === $code ) {
1094 $this->unload( $shallowCode );
1095 }
1096 }
1097 }
1098
1099 /**
1100 * Unload all data
1101 */
1102 public function unloadAll() {
1103 foreach ( $this->initialisedLangs as $lang => $unused ) {
1104 $this->unload( $lang );
1105 }
1106 }
1107
1108 /**
1109 * Disable the storage backend
1110 */
1111 public function disableBackend() {
1112 $this->store = new LCStoreNull;
1113 $this->manualRecache = false;
1114 }
1115 }
1116
1117 /**
1118 * Interface for the persistence layer of LocalisationCache.
1119 *
1120 * The persistence layer is two-level hierarchical cache. The first level
1121 * is the language, the second level is the item or subitem.
1122 *
1123 * Since the data for a whole language is rebuilt in one operation, it needs
1124 * to have a fast and atomic method for deleting or replacing all of the
1125 * current data for a given language. The interface reflects this bulk update
1126 * operation. Callers writing to the cache must first call startWrite(), then
1127 * will call set() a couple of thousand times, then will call finishWrite()
1128 * to commit the operation. When finishWrite() is called, the cache is
1129 * expected to delete all data previously stored for that language.
1130 *
1131 * The values stored are PHP variables suitable for serialize(). Implementations
1132 * of LCStore are responsible for serializing and unserializing.
1133 */
1134 interface LCStore {
1135 /**
1136 * Get a value.
1137 * @param string $code Language code
1138 * @param string $key Cache key
1139 */
1140 function get( $code, $key );
1141
1142 /**
1143 * Start a write transaction.
1144 * @param string $code Language code
1145 */
1146 function startWrite( $code );
1147
1148 /**
1149 * Finish a write transaction.
1150 */
1151 function finishWrite();
1152
1153 /**
1154 * Set a key to a given value. startWrite() must be called before this
1155 * is called, and finishWrite() must be called afterwards.
1156 * @param string $key
1157 * @param mixed $value
1158 */
1159 function set( $key, $value );
1160 }
1161
1162 /**
1163 * LCStore implementation which uses the standard DB functions to store data.
1164 * This will work on any MediaWiki installation.
1165 */
1166 class LCStoreDB implements LCStore {
1167 private $currentLang;
1168 private $writesDone = false;
1169
1170 /** @var DatabaseBase */
1171 private $dbw;
1172 /** @var array */
1173 private $batch = array();
1174
1175 private $readOnly = false;
1176
1177 public function get( $code, $key ) {
1178 if ( $this->writesDone ) {
1179 $db = wfGetDB( DB_MASTER );
1180 } else {
1181 $db = wfGetDB( DB_SLAVE );
1182 }
1183 $row = $db->selectRow( 'l10n_cache', array( 'lc_value' ),
1184 array( 'lc_lang' => $code, 'lc_key' => $key ), __METHOD__ );
1185 if ( $row ) {
1186 return unserialize( $db->decodeBlob( $row->lc_value ) );
1187 } else {
1188 return null;
1189 }
1190 }
1191
1192 public function startWrite( $code ) {
1193 if ( $this->readOnly ) {
1194 return;
1195 } elseif ( !$code ) {
1196 throw new MWException( __METHOD__ . ": Invalid language \"$code\"" );
1197 }
1198
1199 $this->dbw = wfGetDB( DB_MASTER );
1200
1201 $this->currentLang = $code;
1202 $this->batch = array();
1203 }
1204
1205 public function finishWrite() {
1206 if ( $this->readOnly ) {
1207 return;
1208 } elseif ( is_null( $this->currentLang ) ) {
1209 throw new MWException( __CLASS__ . ': must call startWrite() before finishWrite()' );
1210 }
1211
1212 $this->dbw->begin( __METHOD__ );
1213 try {
1214 $this->dbw->delete( 'l10n_cache',
1215 array( 'lc_lang' => $this->currentLang ), __METHOD__ );
1216 foreach ( array_chunk( $this->batch, 500 ) as $rows ) {
1217 $this->dbw->insert( 'l10n_cache', $rows, __METHOD__ );
1218 }
1219 $this->writesDone = true;
1220 } catch ( DBQueryError $e ) {
1221 if ( $this->dbw->wasReadOnlyError() ) {
1222 $this->readOnly = true; // just avoid site down time
1223 } else {
1224 throw $e;
1225 }
1226 }
1227 $this->dbw->commit( __METHOD__ );
1228
1229 $this->currentLang = null;
1230 $this->batch = array();
1231 }
1232
1233 public function set( $key, $value ) {
1234 if ( $this->readOnly ) {
1235 return;
1236 } elseif ( is_null( $this->currentLang ) ) {
1237 throw new MWException( __CLASS__ . ': must call startWrite() before set()' );
1238 }
1239
1240 $this->batch[] = array(
1241 'lc_lang' => $this->currentLang,
1242 'lc_key' => $key,
1243 'lc_value' => $this->dbw->encodeBlob( serialize( $value ) ) );
1244 }
1245 }
1246
1247 /**
1248 * LCStore implementation which stores data as a collection of CDB files in the
1249 * directory given by $wgCacheDirectory. If $wgCacheDirectory is not set, this
1250 * will throw an exception.
1251 *
1252 * Profiling indicates that on Linux, this implementation outperforms MySQL if
1253 * the directory is on a local filesystem and there is ample kernel cache
1254 * space. The performance advantage is greater when the DBA extension is
1255 * available than it is with the PHP port.
1256 *
1257 * See Cdb.php and http://cr.yp.to/cdb.html
1258 */
1259 class LCStoreCDB implements LCStore {
1260 /** @var CdbReader[] */
1261 private $readers;
1262
1263 /** @var CdbWriter */
1264 private $writer;
1265
1266 /** @var string Current language code */
1267 private $currentLang;
1268
1269 /** @var bool|string Cache directory. False if not set */
1270 private $directory;
1271
1272 function __construct( $conf = array() ) {
1273 global $wgCacheDirectory;
1274
1275 if ( isset( $conf['directory'] ) ) {
1276 $this->directory = $conf['directory'];
1277 } else {
1278 $this->directory = $wgCacheDirectory;
1279 }
1280 }
1281
1282 public function get( $code, $key ) {
1283 if ( !isset( $this->readers[$code] ) ) {
1284 $fileName = $this->getFileName( $code );
1285
1286 $this->readers[$code] = false;
1287 if ( file_exists( $fileName ) ) {
1288 try {
1289 $this->readers[$code] = CdbReader::open( $fileName );
1290 } catch ( CdbException $e ) {
1291 wfDebug( __METHOD__ . ": unable to open cdb file for reading\n" );
1292 }
1293 }
1294 }
1295
1296 if ( !$this->readers[$code] ) {
1297 return null;
1298 } else {
1299 $value = false;
1300 try {
1301 $value = $this->readers[$code]->get( $key );
1302 } catch ( CdbException $e ) {
1303 wfDebug( __METHOD__ . ": CdbException caught, error message was "
1304 . $e->getMessage() . "\n" );
1305 }
1306 if ( $value === false ) {
1307 return null;
1308 }
1309
1310 return unserialize( $value );
1311 }
1312 }
1313
1314 public function startWrite( $code ) {
1315 if ( !file_exists( $this->directory ) ) {
1316 if ( !wfMkdirParents( $this->directory, null, __METHOD__ ) ) {
1317 throw new MWException( "Unable to create the localisation store " .
1318 "directory \"{$this->directory}\"" );
1319 }
1320 }
1321
1322 // Close reader to stop permission errors on write
1323 if ( !empty( $this->readers[$code] ) ) {
1324 $this->readers[$code]->close();
1325 }
1326
1327 try {
1328 $this->writer = CdbWriter::open( $this->getFileName( $code ) );
1329 } catch ( CdbException $e ) {
1330 throw new MWException( $e->getMessage() );
1331 }
1332 $this->currentLang = $code;
1333 }
1334
1335 public function finishWrite() {
1336 // Close the writer
1337 try {
1338 $this->writer->close();
1339 } catch ( CdbException $e ) {
1340 throw new MWException( $e->getMessage() );
1341 }
1342 $this->writer = null;
1343 unset( $this->readers[$this->currentLang] );
1344 $this->currentLang = null;
1345 }
1346
1347 public function set( $key, $value ) {
1348 if ( is_null( $this->writer ) ) {
1349 throw new MWException( __CLASS__ . ': must call startWrite() before calling set()' );
1350 }
1351 try {
1352 $this->writer->set( $key, serialize( $value ) );
1353 } catch ( CdbException $e ) {
1354 throw new MWException( $e->getMessage() );
1355 }
1356 }
1357
1358 protected function getFileName( $code ) {
1359 if ( strval( $code ) === '' || strpos( $code, '/' ) !== false ) {
1360 throw new MWException( __METHOD__ . ": Invalid language \"$code\"" );
1361 }
1362
1363 return "{$this->directory}/l10n_cache-$code.cdb";
1364 }
1365 }
1366
1367 /**
1368 * Null store backend, used to avoid DB errors during install
1369 */
1370 class LCStoreNull implements LCStore {
1371 public function get( $code, $key ) {
1372 return null;
1373 }
1374
1375 public function startWrite( $code ) {
1376 }
1377
1378 public function finishWrite() {
1379 }
1380
1381 public function set( $key, $value ) {
1382 }
1383 }
1384
1385 /**
1386 * A localisation cache optimised for loading large amounts of data for many
1387 * languages. Used by rebuildLocalisationCache.php.
1388 */
1389 class LocalisationCacheBulkLoad extends LocalisationCache {
1390 /**
1391 * A cache of the contents of data files.
1392 * Core files are serialized to avoid using ~1GB of RAM during a recache.
1393 */
1394 private $fileCache = array();
1395
1396 /**
1397 * Most recently used languages. Uses the linked-list aspect of PHP hashtables
1398 * to keep the most recently used language codes at the end of the array, and
1399 * the language codes that are ready to be deleted at the beginning.
1400 */
1401 private $mruLangs = array();
1402
1403 /**
1404 * Maximum number of languages that may be loaded into $this->data
1405 */
1406 private $maxLoadedLangs = 10;
1407
1408 /**
1409 * @param string $fileName
1410 * @param string $fileType
1411 * @return array|mixed
1412 */
1413 protected function readPHPFile( $fileName, $fileType ) {
1414 $serialize = $fileType === 'core';
1415 if ( !isset( $this->fileCache[$fileName][$fileType] ) ) {
1416 $data = parent::readPHPFile( $fileName, $fileType );
1417
1418 if ( $serialize ) {
1419 $encData = serialize( $data );
1420 } else {
1421 $encData = $data;
1422 }
1423
1424 $this->fileCache[$fileName][$fileType] = $encData;
1425
1426 return $data;
1427 } elseif ( $serialize ) {
1428 return unserialize( $this->fileCache[$fileName][$fileType] );
1429 } else {
1430 return $this->fileCache[$fileName][$fileType];
1431 }
1432 }
1433
1434 /**
1435 * @param string $code
1436 * @param string $key
1437 * @return mixed
1438 */
1439 public function getItem( $code, $key ) {
1440 unset( $this->mruLangs[$code] );
1441 $this->mruLangs[$code] = true;
1442
1443 return parent::getItem( $code, $key );
1444 }
1445
1446 /**
1447 * @param string $code
1448 * @param string $key
1449 * @param string $subkey
1450 * @return mixed
1451 */
1452 public function getSubitem( $code, $key, $subkey ) {
1453 unset( $this->mruLangs[$code] );
1454 $this->mruLangs[$code] = true;
1455
1456 return parent::getSubitem( $code, $key, $subkey );
1457 }
1458
1459 /**
1460 * @param string $code
1461 */
1462 public function recache( $code ) {
1463 parent::recache( $code );
1464 unset( $this->mruLangs[$code] );
1465 $this->mruLangs[$code] = true;
1466 $this->trimCache();
1467 }
1468
1469 /**
1470 * @param string $code
1471 */
1472 public function unload( $code ) {
1473 unset( $this->mruLangs[$code] );
1474 parent::unload( $code );
1475 }
1476
1477 /**
1478 * Unload cached languages until there are less than $this->maxLoadedLangs
1479 */
1480 protected function trimCache() {
1481 while ( count( $this->data ) > $this->maxLoadedLangs && count( $this->mruLangs ) ) {
1482 reset( $this->mruLangs );
1483 $code = key( $this->mruLangs );
1484 wfDebug( __METHOD__ . ": unloading $code\n" );
1485 $this->unload( $code );
1486 }
1487 }
1488 }