Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
Total | |
0.00% |
0 / 190 |
|
0.00% |
0 / 9 |
CRAP | |
0.00% |
0 / 1 |
ExternalMessageSourceStateComparator | |
0.00% |
0 / 190 |
|
0.00% |
0 / 9 |
2756 | |
0.00% |
0 / 1 |
__construct | |
0.00% |
0 / 3 |
|
0.00% |
0 / 1 |
2 | |||
processGroup | |
0.00% |
0 / 8 |
|
0.00% |
0 / 1 |
6 | |||
processLanguage | |
0.00% |
0 / 6 |
|
0.00% |
0 / 1 |
12 | |||
addMessageUpdateChanges | |
0.00% |
0 / 61 |
|
0.00% |
0 / 1 |
420 | |||
checkNonSourceAdditionsForRename | |
0.00% |
0 / 41 |
|
0.00% |
0 / 1 |
56 | |||
findAndMarkSourceRenames | |
0.00% |
0 / 14 |
|
0.00% |
0 / 1 |
42 | |||
addNonSourceRenames | |
0.00% |
0 / 12 |
|
0.00% |
0 / 1 |
2 | |||
matchRenames | |
0.00% |
0 / 15 |
|
0.00% |
0 / 1 |
20 | |||
hasCacheEntry | |
0.00% |
0 / 30 |
|
0.00% |
0 / 1 |
72 |
1 | <?php |
2 | declare( strict_types = 1 ); |
3 | |
4 | namespace MediaWiki\Extension\Translate\Synchronization; |
5 | |
6 | use FileBasedMessageGroup; |
7 | use MediaWiki\Extension\Translate\MessageGroupProcessing\MessageGroupCache; |
8 | use MediaWiki\Extension\Translate\MessageLoading\Message; |
9 | use MediaWiki\Extension\Translate\MessageLoading\MessageCollection; |
10 | use MediaWiki\Extension\Translate\MessageSync\MessageSourceChange; |
11 | use MediaWiki\Extension\Translate\Utilities\StringComparators\StringComparator; |
12 | use MediaWiki\Extension\Translate\Utilities\Utilities; |
13 | use MediaWiki\Logger\LoggerFactory; |
14 | use MediaWiki\Page\PageStore; |
15 | use MediaWiki\Revision\RevisionLookup; |
16 | use MediaWiki\Utils\MWTimestamp; |
17 | use RuntimeException; |
18 | |
19 | /** |
20 | * Finds external changes for file based message groups. |
21 | * |
22 | * @author Niklas Laxström |
23 | * @license GPL-2.0-or-later |
24 | */ |
25 | class ExternalMessageSourceStateComparator { |
26 | private StringComparator $stringComparator; |
27 | private RevisionLookup $revisionLookup; |
28 | private PageStore $pageStore; |
29 | |
30 | public function __construct( |
31 | StringComparator $stringComparator, |
32 | RevisionLookup $revisionLookup, |
33 | PageStore $pageStore |
34 | ) { |
35 | $this->stringComparator = $stringComparator; |
36 | $this->revisionLookup = $revisionLookup; |
37 | $this->pageStore = $pageStore; |
38 | } |
39 | |
40 | /** |
41 | * Finds modifications in external sources compared to wiki state. |
42 | * |
43 | * The MessageSourceChange object returned stores the following about each modification, |
44 | * - First level of classification is the language code |
45 | * - Second level of classification is the type of modification, |
46 | * - addition (new message in the file) |
47 | * - deletion (message in wiki not present in the file) |
48 | * - change (difference in content) |
49 | * - rename (message key is modified) |
50 | * - Third level is a list of modifications |
51 | * - For each modification, the following is saved, |
52 | * - key (the message key) |
53 | * - content (the message content in external source, null for deletions) |
54 | * - matched_to (present in case of renames, key of the matched message) |
55 | * - similarity (present in case of renames, similarity % with the matched message) |
56 | * - previous_state (present in case of renames, state of the message before rename) |
57 | */ |
58 | public function processGroup( FileBasedMessageGroup $group ): MessageSourceChange { |
59 | $changes = new MessageSourceChange(); |
60 | $languages = $group->getTranslatableLanguages() ?? Utilities::getLanguageNames( 'en' ); |
61 | |
62 | // Process the source language before others. Source language might not |
63 | // be included in $group->getTranslatableLanguages(). The expected |
64 | // behavior is that source language is always processed. |
65 | $sourceLanguage = $group->getSourceLanguage(); |
66 | $this->processLanguage( $group, $sourceLanguage, $changes ); |
67 | unset( $languages[ $sourceLanguage] ); |
68 | |
69 | foreach ( array_keys( $languages ) as $language ) { |
70 | $this->processLanguage( $group, $language, $changes ); |
71 | } |
72 | |
73 | return $changes; |
74 | } |
75 | |
76 | private function processLanguage( |
77 | FileBasedMessageGroup $group, |
78 | string $language, |
79 | MessageSourceChange $changes |
80 | ): void { |
81 | $cache = $group->getMessageGroupCache( $language ); |
82 | $reason = 0; |
83 | if ( !$cache->isValid( $reason ) ) { |
84 | $this->addMessageUpdateChanges( $group, $language, $changes, $reason, $cache ); |
85 | |
86 | if ( $changes->getModificationsForLanguage( $language ) === [] ) { |
87 | /* Update the cache immediately if file and wiki state match. |
88 | * Otherwise the cache will get outdated compared to file state |
89 | * and will give false positive conflicts later. */ |
90 | $cache->create(); |
91 | } |
92 | } |
93 | } |
94 | |
95 | /** |
96 | * This is the detective novel. We have three sources of information: |
97 | * - current message state in the file |
98 | * - current message state in the wiki |
99 | * - cached message state since cache was last build |
100 | * (usually after export from wiki) |
101 | * |
102 | * Now we must try to guess what in earth has driven the file state and |
103 | * wiki state out of sync. Then we must compile list of events that would |
104 | * bring those to sync. Types of events are addition, deletion, (content) |
105 | * change and key renames. After that the list of events are stored for |
106 | * later processing of a translation administrator, who can decide what |
107 | * actions to take on those events to bring the state more or less in sync. |
108 | * |
109 | * @throws RuntimeException |
110 | */ |
111 | protected function addMessageUpdateChanges( |
112 | FileBasedMessageGroup $group, |
113 | string $language, |
114 | MessageSourceChange $changes, |
115 | int $reason, |
116 | MessageGroupCache $cache |
117 | ): void { |
118 | // initCollection returns empty list before first import |
119 | $wiki = $group->initCollection( $language ); |
120 | $wiki->filter( 'hastranslation', false ); |
121 | $wiki->loadTranslations(); |
122 | $wikiKeys = $wiki->getMessageKeys(); |
123 | |
124 | $sourceLanguage = $group->getSourceLanguage(); |
125 | // By-pass cached message definitions |
126 | $ffs = $group->getFFS(); |
127 | if ( $language === $sourceLanguage && !$ffs->exists( $language ) ) { |
128 | $path = $group->getSourceFilePath( $language ); |
129 | throw new RuntimeException( "Source message file for {$group->getId()} does not exist: $path" ); |
130 | } |
131 | |
132 | $file = $ffs->read( $language ); |
133 | |
134 | // Does not exist |
135 | if ( $file === false ) { |
136 | return; |
137 | } |
138 | |
139 | // Something went wrong |
140 | if ( !isset( $file['MESSAGES'] ) ) { |
141 | $id = $group->getId(); |
142 | $ffsClass = get_class( $ffs ); |
143 | |
144 | error_log( "$id has an FFS ($ffsClass) - it didn't return cake for $language" ); |
145 | |
146 | return; |
147 | } |
148 | |
149 | $fileKeys = array_keys( $file['MESSAGES'] ); |
150 | |
151 | $common = array_intersect( $fileKeys, $wikiKeys ); |
152 | |
153 | $supportsFuzzy = $ffs->supportsFuzzy(); |
154 | $changesToRemove = []; |
155 | |
156 | foreach ( $common as $key ) { |
157 | $sourceContent = $file['MESSAGES'][$key]; |
158 | /** @var Message $wikiMessage */ |
159 | $wikiMessage = $wiki[$key]; |
160 | $wikiContent = $wikiMessage->translation(); |
161 | |
162 | // @todo: Fuzzy checking can also be moved to $ffs->isContentEqual(); |
163 | // If FFS doesn't support it, ignore fuzziness as difference |
164 | $wikiContent = str_replace( TRANSLATE_FUZZY, '', $wikiContent ); |
165 | |
166 | // But if it does, ensure we have exactly one fuzzy marker prefixed |
167 | if ( $supportsFuzzy === 'yes' && $wikiMessage->hasTag( 'fuzzy' ) ) { |
168 | $wikiContent = TRANSLATE_FUZZY . $wikiContent; |
169 | } |
170 | |
171 | if ( $ffs->isContentEqual( $sourceContent, $wikiContent ) ) { |
172 | // File and wiki stage agree, nothing to do |
173 | continue; |
174 | } |
175 | |
176 | // Check against interim cache to see whether we have changes |
177 | // in the wiki, in the file or both. |
178 | |
179 | if ( $reason !== MessageGroupCache::NO_CACHE ) { |
180 | $cacheContent = $cache->get( $key ); |
181 | |
182 | /* We want to ignore the following situations: |
183 | * 1. The string in the wiki has been changed since the last export. |
184 | * Hence we check that source === cache && cache !== wiki |
185 | * 2. Missing cache entry due to the string being translated on translatewiki.net, |
186 | * exported and then being updated on translatewiki.net again. |
187 | */ |
188 | if ( |
189 | $this->hasCacheEntry( $cache, $wiki, $key ) && |
190 | !$ffs->isContentEqual( $wikiContent, $cacheContent ) && |
191 | $ffs->isContentEqual( $sourceContent, $cacheContent ) |
192 | ) { |
193 | continue; |
194 | } |
195 | } |
196 | |
197 | if ( $language !== $sourceLanguage ) { |
198 | // Assuming that this is the old key, lets check if it has a corresponding |
199 | // rename in the source language. The key of the matching message will be |
200 | // the new renamed key. |
201 | $renameMsg = $changes->getMatchedMessage( $sourceLanguage, $key ); |
202 | if ( $renameMsg !== null ) { |
203 | // Rename present in source language but this message has a content change |
204 | // with the OLD key in a non-source language. We will not process this |
205 | // here but add it as a rename instead. This way, the key will be renamed |
206 | // and then the content updated. |
207 | $this->addNonSourceRenames( |
208 | $changes, $key, $renameMsg['key'], $sourceContent, $wikiContent, $language |
209 | ); |
210 | $changesToRemove[] = $key; |
211 | continue; |
212 | } |
213 | } |
214 | $changes->addChange( $language, $key, $sourceContent ); |
215 | } |
216 | |
217 | $changes->removeChanges( $language, $changesToRemove ); |
218 | |
219 | $added = array_diff( $fileKeys, $wikiKeys ); |
220 | foreach ( $added as $key ) { |
221 | $sourceContent = $file['MESSAGES'][$key]; |
222 | $changes->addAddition( $language, $key, $sourceContent ); |
223 | } |
224 | |
225 | /* Should the cache not exist, don't consider the messages |
226 | * missing from the file as deleted - they probably aren't |
227 | * yet exported. For example new language translations are |
228 | * exported the first time. */ |
229 | if ( $reason !== MessageGroupCache::NO_CACHE ) { |
230 | $deleted = array_diff( $wikiKeys, $fileKeys ); |
231 | foreach ( $deleted as $key ) { |
232 | if ( $cache->get( $key ) === false ) { |
233 | /* This message has never existed in the cache, so it |
234 | * must be a newly made in the wiki. */ |
235 | continue; |
236 | } |
237 | $changes->addDeletion( $language, $key, $wiki[$key]->translation() ); |
238 | } |
239 | } |
240 | |
241 | if ( $language === $sourceLanguage ) { |
242 | $this->findAndMarkSourceRenames( $changes, $language ); |
243 | } else { |
244 | // Non source language |
245 | $this->checkNonSourceAdditionsForRename( |
246 | $changes, $sourceLanguage, $language, $wiki, $wikiKeys |
247 | ); |
248 | } |
249 | } |
250 | |
251 | /** |
252 | * For non source languages, we look at additions and see if they have been |
253 | * added as renames in the source language. |
254 | * @param MessageSourceChange $changes |
255 | * @param string $sourceLanguage |
256 | * @param string $targetLanguage |
257 | * @param MessageCollection $wiki |
258 | * @param string[] $wikiKeys |
259 | */ |
260 | private function checkNonSourceAdditionsForRename( |
261 | MessageSourceChange $changes, |
262 | string $sourceLanguage, |
263 | string $targetLanguage, |
264 | MessageCollection $wiki, |
265 | array $wikiKeys |
266 | ): void { |
267 | $additions = $changes->getAdditions( $targetLanguage ); |
268 | if ( $additions === [] ) { |
269 | return; |
270 | } |
271 | |
272 | $additionsToRemove = []; |
273 | $deletionsToRemove = []; |
274 | foreach ( $additions as $addedMsg ) { |
275 | $addedMsgKey = $addedMsg['key']; |
276 | |
277 | // Check if this key is renamed in source. |
278 | $renamedSourceMsg = $changes->findMessage( |
279 | $sourceLanguage, $addedMsgKey, [ MessageSourceChange::RENAME ] |
280 | ); |
281 | |
282 | if ( $renamedSourceMsg === null ) { |
283 | continue; |
284 | } |
285 | |
286 | // Since this key is new, and is present in the renames for the source language, |
287 | // we will add it as a rename. |
288 | $deletedSource = $changes->getMatchedMessage( $sourceLanguage, $renamedSourceMsg['key'] ); |
289 | if ( $deletedSource === null ) { |
290 | continue; |
291 | } |
292 | $deletedMsgKey = $deletedSource['key']; |
293 | $deletedMsg = $changes->findMessage( |
294 | $targetLanguage, $deletedMsgKey, [ MessageSourceChange::DELETION ] |
295 | ); |
296 | |
297 | // Sometimes when the cache does not have the translations, the deleted message |
298 | // is not added in the translations. It is also possible that for this non-source |
299 | // language the key has not been removed. |
300 | if ( $deletedMsg === null ) { |
301 | $content = ''; |
302 | if ( in_array( $deletedMsgKey, $wikiKeys ) ) { |
303 | $content = $wiki[ $deletedMsgKey ]->translation(); |
304 | } |
305 | $deletedMsg = [ |
306 | 'key' => $deletedMsgKey, |
307 | 'content' => $content |
308 | ]; |
309 | } |
310 | |
311 | $similarityPercent = $this->stringComparator->getSimilarity( |
312 | $addedMsg['content'], $deletedMsg['content'] |
313 | ); |
314 | |
315 | $changes->addRename( $targetLanguage, [ |
316 | 'key' => $addedMsgKey, |
317 | 'content' => $addedMsg['content'] |
318 | ], [ |
319 | 'key' => $deletedMsgKey, |
320 | 'content' => $deletedMsg['content'] |
321 | ], $similarityPercent ); |
322 | |
323 | $deletionsToRemove[] = $deletedMsgKey; |
324 | $additionsToRemove[] = $addedMsgKey; |
325 | } |
326 | |
327 | $changes->removeAdditions( $targetLanguage, $additionsToRemove ); |
328 | $changes->removeDeletions( $targetLanguage, $deletionsToRemove ); |
329 | } |
330 | |
331 | /** |
332 | * Check for renames and add them to the changes. To identify renames we need to |
333 | * compare the contents of the added messages with the deleted ones and identify |
334 | * messages that match. |
335 | */ |
336 | private function findAndMarkSourceRenames( MessageSourceChange $changes, string $sourceLanguage ): void { |
337 | // Now check for renames. To identify renames we need to compare |
338 | // the contents of the added messages with the deleted ones and |
339 | // identify messages that match. |
340 | $deletions = $changes->getDeletions( $sourceLanguage ); |
341 | $additions = $changes->getAdditions( $sourceLanguage ); |
342 | if ( $deletions === [] || $additions === [] ) { |
343 | return; |
344 | } |
345 | |
346 | // This array contains a dictionary with matching renames in the following structure - |
347 | // [ A1|D1 => 1.0, A1|D2 => 0.95, A2|D1 => 0.95 ] |
348 | $potentialRenames = []; |
349 | foreach ( $additions as $addedMsg ) { |
350 | $addedMsgKey = $addedMsg['key']; |
351 | |
352 | foreach ( $deletions as $deletedMsg ) { |
353 | $similarityPercent = $this->stringComparator->getSimilarity( |
354 | $addedMsg['content'], $deletedMsg['content'] |
355 | ); |
356 | |
357 | if ( $changes->areStringsSimilar( $similarityPercent ) ) { |
358 | $potentialRenames[ $addedMsgKey . '|' . $deletedMsg['key'] ] = $similarityPercent; |
359 | } |
360 | } |
361 | } |
362 | |
363 | $this->matchRenames( $changes, $potentialRenames, $sourceLanguage ); |
364 | } |
365 | |
366 | /** Adds non source language renames to the list of changes */ |
367 | private function addNonSourceRenames( |
368 | MessageSourceChange $changes, |
369 | string $key, |
370 | string $renameKey, |
371 | string $sourceContent, |
372 | string $wikiContent, |
373 | string $language |
374 | ): void { |
375 | $addedMsg = [ |
376 | 'key' => $renameKey, |
377 | 'content' => $sourceContent |
378 | ]; |
379 | |
380 | $removedMsg = [ |
381 | 'key' => $key, |
382 | 'content' => $wikiContent |
383 | ]; |
384 | |
385 | $similarityPercent = $this->stringComparator->getSimilarity( |
386 | $sourceContent, $wikiContent |
387 | ); |
388 | $changes->addRename( $language, $addedMsg, $removedMsg, $similarityPercent ); |
389 | } |
390 | |
391 | /** |
392 | * Identifies which added message to be associated with the deleted message based on |
393 | * similarity percentage. |
394 | * |
395 | * We sort the $trackRename array on the similarity percentage and then start adding the |
396 | * messages as renames. |
397 | */ |
398 | private function matchRenames( MessageSourceChange $changes, array $trackRename, string $language ): void { |
399 | arsort( $trackRename, SORT_NUMERIC ); |
400 | |
401 | $alreadyRenamed = $additionsToRemove = $deletionsToRemove = []; |
402 | foreach ( $trackRename as $key => $similarityPercent ) { |
403 | [ $addKey, $deleteKey ] = explode( '|', $key, 2 ); |
404 | if ( isset( $alreadyRenamed[ $addKey ] ) || isset( $alreadyRenamed[ $deleteKey ] ) ) { |
405 | // Already mapped with another name. |
406 | continue; |
407 | } |
408 | |
409 | // Using key should be faster than saving values and searching for them in the array. |
410 | $alreadyRenamed[ $addKey ] = 1; |
411 | $alreadyRenamed[ $deleteKey ] = 1; |
412 | |
413 | $addMsg = $changes->findMessage( $language, $addKey, [ MessageSourceChange::ADDITION ] ); |
414 | $deleteMsg = $changes->findMessage( $language, $deleteKey, [ MessageSourceChange::DELETION ] ); |
415 | |
416 | $changes->addRename( $language, $addMsg, $deleteMsg, $similarityPercent ); |
417 | |
418 | // @phan-suppress-next-line PhanTypeArraySuspiciousNullable |
419 | $additionsToRemove[] = $addMsg['key']; |
420 | // @phan-suppress-next-line PhanTypeArraySuspiciousNullable |
421 | $deletionsToRemove[] = $deleteMsg['key']; |
422 | } |
423 | |
424 | $changes->removeAdditions( $language, $additionsToRemove ); |
425 | $changes->removeDeletions( $language, $deletionsToRemove ); |
426 | } |
427 | |
428 | /** |
429 | * Checks if the cache has an entry for the given key |
430 | * @return bool True if entry is present, false if entry is not present but that is the expected behavior |
431 | * @throws RuntimeException If the cache should have an entry but is unexpectedly missing |
432 | */ |
433 | private function hasCacheEntry( |
434 | MessageGroupCache $cache, |
435 | MessageCollection $collection, |
436 | string $messageKey |
437 | ): bool { |
438 | $cacheContent = $cache->get( $messageKey ); |
439 | if ( $cacheContent !== false ) { |
440 | return true; |
441 | } |
442 | |
443 | $cacheUpdateTime = $cache->getUpdateTimestamp(); |
444 | $cacheUpdateTime = $cacheUpdateTime !== false ? MWTimestamp::convert( TS_MW, $cacheUpdateTime ) : false; |
445 | |
446 | $pageIdentity = $this->pageStore->getPageForLink( $collection->keys()[ $messageKey ] ); |
447 | $oldestRevision = $this->revisionLookup->getFirstRevision( $pageIdentity ); |
448 | $latestRevision = $this->revisionLookup->getRevisionByTitle( $pageIdentity ); |
449 | |
450 | $logger = LoggerFactory::getInstance( 'Translate' ); |
451 | // Here we are checking for the following: |
452 | // 1. New translation was added for a message on translatewiki.net |
453 | // 2. Translation was exported |
454 | // 3. Translation was updated on translatewiki.net |
455 | // In this case the cache does not have the message |
456 | if ( |
457 | $cacheUpdateTime !== false && |
458 | ( $oldestRevision && $oldestRevision->getTimestamp() < $cacheUpdateTime ) && |
459 | ( $latestRevision && $cacheUpdateTime < $latestRevision->getTimestamp() ) |
460 | ) { |
461 | $logger->info( |
462 | 'Expected cache miss for {messageKey} in language: {language}. Cache update time: {cacheUpdateTime}', |
463 | [ |
464 | 'messageKey' => $messageKey, |
465 | 'language' => $collection->getLanguage(), |
466 | 'cacheUpdateTime' => $cacheUpdateTime |
467 | ] |
468 | ); |
469 | return false; |
470 | } |
471 | |
472 | $logger->warning( |
473 | 'Unexpected cache miss for {messageKey} in language: {language}. Cache update time: {cacheUpdateTime}', |
474 | [ |
475 | 'messageKey' => $messageKey, |
476 | 'language' => $collection->getLanguage(), |
477 | 'cacheUpdateTime' => $cacheUpdateTime |
478 | ] |
479 | ); |
480 | return false; |
481 | } |
482 | |
483 | } |