Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
86.96% |
120 / 138 |
|
66.67% |
12 / 18 |
CRAP | |
0.00% |
0 / 1 |
| SegmentPageFactory | |
86.96% |
120 / 138 |
|
66.67% |
12 / 18 |
45.91 | |
0.00% |
0 / 1 |
| __construct | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| setUseSegmentsCache | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| setUseRevisionPropertiesCache | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| setSegmenter | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| setSegmenterByLanguage | |
0.00% |
0 / 1 |
|
0.00% |
0 / 1 |
2 | |||
| setRemoveTags | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| setSegmentBreakingTags | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| setContextSource | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| setRevisionStore | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| setHttpRequestFactory | |
0.00% |
0 / 2 |
|
0.00% |
0 / 1 |
2 | |||
| setConsumerUrl | |
0.00% |
0 / 2 |
|
0.00% |
0 / 1 |
2 | |||
| setRequirePageRevisionProperties | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| pageProviderFactory | |
44.44% |
4 / 9 |
|
0.00% |
0 / 1 |
9.29 | |||
| loadPageRevisionProperties | |
77.78% |
7 / 9 |
|
0.00% |
0 / 1 |
4.18 | |||
| segmentPage | |
91.78% |
67 / 73 |
|
0.00% |
0 / 1 |
18.18 | |||
| segmentedPageCacheKeyFactory | |
100.00% |
12 / 12 |
|
100.00% |
1 / 1 |
1 | |||
| pageRevisionPropertiesCacheKeyFactory | |
100.00% |
10 / 10 |
|
100.00% |
1 / 1 |
1 | |||
| cleanHtmlDom | |
100.00% |
2 / 2 |
|
100.00% |
1 / 1 |
1 | |||
| 1 | <?php |
| 2 | |
| 3 | namespace MediaWiki\Wikispeech\Segment; |
| 4 | |
| 5 | /** |
| 6 | * @file |
| 7 | * @ingroup Extensions |
| 8 | * @license GPL-2.0-or-later |
| 9 | */ |
| 10 | |
| 11 | use ConfigFactory; |
| 12 | use IContextSource; |
| 13 | use InvalidArgumentException; |
| 14 | use LogicException; |
| 15 | use MediaWiki\Http\HttpRequestFactory; |
| 16 | use MediaWiki\Revision\RevisionStore; |
| 17 | use MWException; |
| 18 | use Title; |
| 19 | use WANObjectCache; |
| 20 | |
| 21 | /** |
| 22 | * @since 0.1.10 |
| 23 | */ |
| 24 | class SegmentPageFactory { |
| 25 | |
| 26 | /** @var WANObjectCache */ |
| 27 | private $cache; |
| 28 | |
| 29 | /** @var ConfigFactory */ |
| 30 | private $configFactory; |
| 31 | |
| 32 | /** |
| 33 | * Whether or not to use cache for segmentation |
| 34 | * @var bool |
| 35 | */ |
| 36 | private $useSegmentsCache = true; |
| 37 | |
| 38 | /** |
| 39 | * Whether or not to use cache for page revision properties, |
| 40 | * i.e. to retrieve page id and title when only supplying a revision id. |
| 41 | * |
| 42 | * Not turning this on in a situation where consumerUrl is set |
| 43 | * will cause one extra http request to the remote wiki |
| 44 | * in order to lookup pageId, which is required to create or find |
| 45 | * utterances in UtteranceStore. |
| 46 | * |
| 47 | * @var bool |
| 48 | */ |
| 49 | private $useRevisionPropertiesCache = false; |
| 50 | |
| 51 | /** |
| 52 | * Will default to an instance of {@link StandardSegmenter} if not set. |
| 53 | * @var Segmenter|null |
| 54 | */ |
| 55 | private $segmenter = null; |
| 56 | |
| 57 | /** |
| 58 | * Will default to config setting WikispeechRemoveTags if not set. |
| 59 | * @var string[]|null |
| 60 | */ |
| 61 | private $removeTags = null; |
| 62 | |
| 63 | /** |
| 64 | * Will default to config setting WikispeechSegmentBreakingTags if not set. |
| 65 | * @var string[]|null |
| 66 | */ |
| 67 | private $segmentBreakingTags = null; |
| 68 | |
| 69 | /** |
| 70 | * Required only when providing page content from a local wiki. |
| 71 | * @var IContextSource |
| 72 | */ |
| 73 | private $contextSource = null; |
| 74 | |
| 75 | /** |
| 76 | * Required only when providing page content from a local wiki. |
| 77 | * @var RevisionStore|null |
| 78 | */ |
| 79 | private $revisionStore = null; |
| 80 | |
| 81 | /** |
| 82 | * Required only when providing page content from a remote wiki. |
| 83 | * @var HttpRequestFactory|null |
| 84 | */ |
| 85 | private $httpRequestFactory = null; |
| 86 | |
| 87 | /** |
| 88 | * Required only when providing page content from a remote wiki. |
| 89 | * @var string|null |
| 90 | */ |
| 91 | private $consumerUrl = null; |
| 92 | |
| 93 | /** |
| 94 | * If true, page id and title is always retrieved from page provider or cache |
| 95 | * and be made available in response from segmentPage. |
| 96 | * |
| 97 | * @var bool |
| 98 | */ |
| 99 | private $requirePageRevisionProperties = false; |
| 100 | |
| 101 | /** |
| 102 | * @since 0.1.10 |
| 103 | * @param WANObjectCache $cache |
| 104 | * @param ConfigFactory $configFactory |
| 105 | */ |
| 106 | public function __construct( |
| 107 | WANObjectCache $cache, |
| 108 | ConfigFactory $configFactory |
| 109 | ) { |
| 110 | $this->cache = $cache; |
| 111 | $this->configFactory = $configFactory; |
| 112 | } |
| 113 | |
| 114 | /** |
| 115 | * @see SegmentPageFactory::$useSegmentsCache |
| 116 | * @since 0.1.10 |
| 117 | * @param bool $useSegmentsCache |
| 118 | * @return SegmentPageFactory $this |
| 119 | */ |
| 120 | public function setUseSegmentsCache( |
| 121 | bool $useSegmentsCache |
| 122 | ): SegmentPageFactory { |
| 123 | $this->useSegmentsCache = $useSegmentsCache; |
| 124 | return $this; |
| 125 | } |
| 126 | |
| 127 | /** |
| 128 | * @see SegmentPageFactory::$useRevisionPropertiesCache |
| 129 | * @since 0.1.10 |
| 130 | * @param bool $useRevisionPropertiesCache |
| 131 | * @return SegmentPageFactory $this |
| 132 | */ |
| 133 | public function setUseRevisionPropertiesCache( |
| 134 | bool $useRevisionPropertiesCache |
| 135 | ): SegmentPageFactory { |
| 136 | $this->useRevisionPropertiesCache = $useRevisionPropertiesCache; |
| 137 | return $this; |
| 138 | } |
| 139 | |
| 140 | /** |
| 141 | * @see SegmentPageFactory::$segmenter |
| 142 | * @since 0.1.10 |
| 143 | * @param Segmenter|null $segmenter |
| 144 | * @return SegmentPageFactory $this |
| 145 | */ |
| 146 | public function setSegmenter( |
| 147 | ?Segmenter $segmenter |
| 148 | ): SegmentPageFactory { |
| 149 | $this->segmenter = $segmenter; |
| 150 | return $this; |
| 151 | } |
| 152 | |
| 153 | /** |
| 154 | * @see SegmentPageFactory::$setSegmenter |
| 155 | * @since 0.1.10 |
| 156 | * @param string $language |
| 157 | * @return SegmentPageFactory $this |
| 158 | */ |
| 159 | public function setSegmenterByLanguage( |
| 160 | string $language |
| 161 | ): SegmentPageFactory { |
| 162 | // @todo lookup segmenter by language |
| 163 | return $this->setSegmenter( new StandardSegmenter() ); |
| 164 | } |
| 165 | |
| 166 | /** |
| 167 | * @see SegmentPageFactory::$removeTags |
| 168 | * @since 0.1.10 |
| 169 | * @param string[]|null $removeTags |
| 170 | * @return SegmentPageFactory $this |
| 171 | */ |
| 172 | public function setRemoveTags( |
| 173 | ?array $removeTags |
| 174 | ): SegmentPageFactory { |
| 175 | $this->removeTags = $removeTags; |
| 176 | return $this; |
| 177 | } |
| 178 | |
| 179 | /** |
| 180 | * @see SegmentPageFactory::$segmentBreakingTags |
| 181 | * @since 0.1.10 |
| 182 | * @param string[]|null $segmentBreakingTags |
| 183 | * @return SegmentPageFactory $this |
| 184 | */ |
| 185 | public function setSegmentBreakingTags( |
| 186 | ?array $segmentBreakingTags |
| 187 | ): SegmentPageFactory { |
| 188 | $this->segmentBreakingTags = $segmentBreakingTags; |
| 189 | return $this; |
| 190 | } |
| 191 | |
| 192 | /** |
| 193 | * @see SegmentPageFactory::$contextSource |
| 194 | * @since 0.1.10 |
| 195 | * @param IContextSource|null $contextSource |
| 196 | * @return SegmentPageFactory $this |
| 197 | */ |
| 198 | public function setContextSource( |
| 199 | ?IContextSource $contextSource |
| 200 | ): SegmentPageFactory { |
| 201 | $this->contextSource = $contextSource; |
| 202 | return $this; |
| 203 | } |
| 204 | |
| 205 | /** |
| 206 | * @see SegmentPageFactory::$revisionStore |
| 207 | * @since 0.1.10 |
| 208 | * @param RevisionStore|null $revisionStore |
| 209 | * @return SegmentPageFactory $this |
| 210 | */ |
| 211 | public function setRevisionStore( |
| 212 | ?RevisionStore $revisionStore |
| 213 | ): SegmentPageFactory { |
| 214 | $this->revisionStore = $revisionStore; |
| 215 | return $this; |
| 216 | } |
| 217 | |
| 218 | /** |
| 219 | * @see SegmentPageFactory::$httpRequestFactory |
| 220 | * @since 0.1.10 |
| 221 | * @param HttpRequestFactory|null $httpRequestFactory |
| 222 | * @return SegmentPageFactory $this |
| 223 | */ |
| 224 | public function setHttpRequestFactory( |
| 225 | ?HttpRequestFactory $httpRequestFactory |
| 226 | ): SegmentPageFactory { |
| 227 | $this->httpRequestFactory = $httpRequestFactory; |
| 228 | return $this; |
| 229 | } |
| 230 | |
| 231 | /** |
| 232 | * @see SegmentPageFactory::$consumerUrl |
| 233 | * @since 0.1.10 |
| 234 | * @param string|null $consumerUrl |
| 235 | * @return SegmentPageFactory $this |
| 236 | */ |
| 237 | public function setConsumerUrl( |
| 238 | ?string $consumerUrl |
| 239 | ): SegmentPageFactory { |
| 240 | $this->consumerUrl = $consumerUrl; |
| 241 | return $this; |
| 242 | } |
| 243 | |
| 244 | /** |
| 245 | * If true, page id and title is always retrieved from page provider or cache |
| 246 | * and be made available in response from segmentPage. |
| 247 | * |
| 248 | * @since 0.1.10 |
| 249 | * @param bool $requirePageRevisionProperties |
| 250 | * @return SegmentPageFactory $this |
| 251 | */ |
| 252 | public function setRequirePageRevisionProperties( |
| 253 | bool $requirePageRevisionProperties |
| 254 | ): SegmentPageFactory { |
| 255 | $this->requirePageRevisionProperties = $requirePageRevisionProperties; |
| 256 | return $this; |
| 257 | } |
| 258 | |
| 259 | /** |
| 260 | * @since 0.1.10 |
| 261 | * @return PageProvider |
| 262 | */ |
| 263 | protected function pageProviderFactory(): PageProvider { |
| 264 | if ( $this->consumerUrl ) { |
| 265 | if ( $this->httpRequestFactory === null ) { |
| 266 | throw new LogicException( '$httpRequestFactory is null!' ); |
| 267 | } |
| 268 | return new RemoteWikiPageProvider( $this->consumerUrl, $this->httpRequestFactory ); |
| 269 | } else { |
| 270 | if ( $this->contextSource === null ) { |
| 271 | throw new LogicException( '$contextSource is null!' ); |
| 272 | } |
| 273 | if ( $this->revisionStore === null ) { |
| 274 | throw new LogicException( '$revisionStore is null!' ); |
| 275 | } |
| 276 | return new LocalWikiPageProvider( $this->contextSource, $this->revisionStore ); |
| 277 | } |
| 278 | } |
| 279 | |
| 280 | /** |
| 281 | * Loads revision properties |
| 282 | * from cache (if set to use) |
| 283 | * or via pageProvider (if not using or missing in cache). |
| 284 | * |
| 285 | * @since 0.1.10 |
| 286 | * @param PageProvider $pageProvider |
| 287 | * @param int $revisionId |
| 288 | * @return PageRevisionProperties |
| 289 | */ |
| 290 | protected function loadPageRevisionProperties( |
| 291 | PageProvider $pageProvider, |
| 292 | int $revisionId |
| 293 | ): PageRevisionProperties { |
| 294 | $cacheKey = $this->pageRevisionPropertiesCacheKeyFactory( $pageProvider, $revisionId ); |
| 295 | // Lookup title and page id given the revision id. |
| 296 | if ( $this->useRevisionPropertiesCache ) { |
| 297 | $revisionProperties = $this->cache->get( $cacheKey ); |
| 298 | } else { |
| 299 | $revisionProperties = false; |
| 300 | } |
| 301 | if ( $revisionProperties === false ) { |
| 302 | $revisionProperties = $pageProvider->loadPageRevisionProperties( $revisionId ); |
| 303 | if ( $this->useRevisionPropertiesCache ) { |
| 304 | $this->cache->set( $cacheKey, $revisionProperties ); |
| 305 | } |
| 306 | } |
| 307 | return $revisionProperties; |
| 308 | } |
| 309 | |
| 310 | /** |
| 311 | * Convenience function to build segmenter |
| 312 | * which is immediately invoked to segment page |
| 313 | * after parsing DOM of the HTML |
| 314 | * supplied by a page provider that is constructed from the factory fields, |
| 315 | * possibly loading extra properties required by invoker, |
| 316 | * and possibly bypassing all of above invocation using caches. |
| 317 | * |
| 318 | * $title XOR $revisionId must be provided. |
| 319 | * |
| 320 | * @since 0.1.10 |
| 321 | * @param Title|null $title |
| 322 | * @param int|null $revisionId |
| 323 | * @return SegmentPageResponse |
| 324 | * @throws InvalidArgumentException If $title xor $revisionId is not provided. |
| 325 | */ |
| 326 | public function segmentPage( |
| 327 | ?Title $title, |
| 328 | ?int $revisionId |
| 329 | ): SegmentPageResponse { |
| 330 | $pageProvider = $this->pageProviderFactory(); |
| 331 | |
| 332 | $segmentPageResponse = new SegmentPageResponse(); |
| 333 | |
| 334 | $revisionProperties = null; |
| 335 | |
| 336 | if ( $title === null && $revisionId !== null ) { |
| 337 | // Lookup title and page id given the revision id. |
| 338 | $revisionProperties = $this->loadPageRevisionProperties( $pageProvider, $revisionId ); |
| 339 | $segmentPageResponse->setTitle( $revisionProperties->getTitle() ); |
| 340 | $segmentPageResponse->setPageId( $revisionProperties->getPageId() ); |
| 341 | $segmentPageResponse->setRevisionId( $revisionId ); |
| 342 | } elseif ( $title !== null && $revisionId === null ) { |
| 343 | // Set user supplied title. |
| 344 | $segmentPageResponse->setTitle( $title ); |
| 345 | } else { |
| 346 | throw new InvalidArgumentException( '$title xor $revisionId must be provided.' ); |
| 347 | } |
| 348 | |
| 349 | $config = $this->configFactory->makeConfig( 'wikispeech' ); |
| 350 | $removeTags = $this->removeTags ?? $config->get( 'WikispeechRemoveTags' ); |
| 351 | $segmentBreakingTags = $this->segmentBreakingTags ?? $config->get( 'WikispeechSegmentBreakingTags' ); |
| 352 | |
| 353 | $segmenter = $this->segmenter ?? new StandardSegmenter(); |
| 354 | |
| 355 | if ( $this->useSegmentsCache && $revisionId !== null ) { |
| 356 | $segments = $this->cache->get( |
| 357 | $this->segmentedPageCacheKeyFactory( |
| 358 | $removeTags, |
| 359 | $segmentBreakingTags, |
| 360 | $pageProvider, |
| 361 | $revisionId |
| 362 | ) |
| 363 | ); |
| 364 | } else { |
| 365 | $segments = false; |
| 366 | } |
| 367 | if ( $segments === false ) { |
| 368 | $segmentPageResponseTitle = $segmentPageResponse->getTitle(); |
| 369 | if ( $segmentPageResponseTitle === null ) { |
| 370 | throw new LogicException( 'Title is null!' ); |
| 371 | } |
| 372 | $pageProvider->loadData( $segmentPageResponseTitle ); |
| 373 | |
| 374 | $displayTitle = $pageProvider->getDisplayTitle(); |
| 375 | if ( $displayTitle === null ) { |
| 376 | throw new MWException( 'Display title not loaded!' ); |
| 377 | } |
| 378 | $pageContent = $pageProvider->getPageContent(); |
| 379 | if ( $pageContent === null ) { |
| 380 | throw new MWException( 'Page content not loaded!' ); |
| 381 | } |
| 382 | $providedRevisionId = $pageProvider->getRevisionId(); |
| 383 | if ( $providedRevisionId === null ) { |
| 384 | throw new MWException( 'Revision id not loaded!' ); |
| 385 | } |
| 386 | |
| 387 | if ( $revisionId !== null |
| 388 | && $revisionId !== $providedRevisionId |
| 389 | ) { |
| 390 | throw new OutdatedOrInvalidRevisionException( 'An outdated or invalid revision id was provided' ); |
| 391 | } |
| 392 | |
| 393 | $cleanedText = $this->cleanHtmlDom( |
| 394 | $displayTitle, |
| 395 | $pageContent, |
| 396 | $removeTags, |
| 397 | $segmentBreakingTags |
| 398 | ); |
| 399 | |
| 400 | $segments = new SegmentList( $segmenter->segmentSentences( $cleanedText ) ); |
| 401 | |
| 402 | if ( $this->useSegmentsCache ) { |
| 403 | $this->cache->set( |
| 404 | // use revision as stated by page provider, not as provided by invoker. |
| 405 | $this->segmentedPageCacheKeyFactory( |
| 406 | $removeTags, |
| 407 | $segmentBreakingTags, |
| 408 | $pageProvider, |
| 409 | $providedRevisionId |
| 410 | ), |
| 411 | $segments, |
| 412 | WANObjectCache::TTL_HOUR |
| 413 | ); |
| 414 | } |
| 415 | $segmentPageResponse->setRevisionId( $providedRevisionId ); |
| 416 | } |
| 417 | |
| 418 | $segmentPageResponse->setSegments( $segments ); |
| 419 | |
| 420 | if ( $this->requirePageRevisionProperties && $revisionProperties === null ) { |
| 421 | $segmentPageResponseRevisionId = $segmentPageResponse->getRevisionId(); |
| 422 | if ( $segmentPageResponseRevisionId === null ) { |
| 423 | throw new LogicException( 'Revision id is null!' ); |
| 424 | } |
| 425 | $revisionProperties = $this->loadPageRevisionProperties( |
| 426 | $pageProvider, |
| 427 | $segmentPageResponseRevisionId |
| 428 | ); |
| 429 | $segmentPageResponse->setTitle( $revisionProperties->getTitle() ); |
| 430 | $segmentPageResponse->setPageId( $revisionProperties->getPageId() ); |
| 431 | } |
| 432 | |
| 433 | return $segmentPageResponse; |
| 434 | } |
| 435 | |
| 436 | /** |
| 437 | * @param string[] $removeTags |
| 438 | * @param string[] $segmentBreakingTags |
| 439 | * @param PageProvider $pageProvider |
| 440 | * @param int|null $revisionId |
| 441 | * @return string |
| 442 | */ |
| 443 | private function segmentedPageCacheKeyFactory( |
| 444 | array $removeTags, |
| 445 | array $segmentBreakingTags, |
| 446 | PageProvider $pageProvider, |
| 447 | ?int $revisionId |
| 448 | ): string { |
| 449 | $cacheKeyComponents = [ |
| 450 | get_class( $this ), |
| 451 | get_class( $pageProvider ), |
| 452 | $revisionId, |
| 453 | var_export( $removeTags, true ), |
| 454 | implode( '-', $segmentBreakingTags ), |
| 455 | $pageProvider->getCachedSegmentsKeyComponents() |
| 456 | ]; |
| 457 | return $this->cache->makeKey( |
| 458 | 'Wikispeech.segments', |
| 459 | ...$cacheKeyComponents |
| 460 | ); |
| 461 | } |
| 462 | |
| 463 | /** |
| 464 | * @param PageProvider $pageProvider |
| 465 | * @param int $revisionId |
| 466 | * @return string |
| 467 | */ |
| 468 | private function pageRevisionPropertiesCacheKeyFactory( |
| 469 | PageProvider $pageProvider, |
| 470 | int $revisionId |
| 471 | ): string { |
| 472 | $cacheKeyComponents = [ |
| 473 | get_class( $this ), |
| 474 | get_class( $pageProvider ), |
| 475 | $revisionId, |
| 476 | $pageProvider->getCachedSegmentsKeyComponents() |
| 477 | ]; |
| 478 | return $this->cache->makeKey( |
| 479 | 'Wikispeech.pageRevisionProperties', |
| 480 | ...$cacheKeyComponents |
| 481 | ); |
| 482 | } |
| 483 | |
| 484 | /** |
| 485 | * This method exists due to need for test mocking. |
| 486 | * |
| 487 | * @see Cleaner::cleanHtmlDom() |
| 488 | * @since 0.1.10 |
| 489 | * @param string $displayTitle |
| 490 | * @param string $pageContent |
| 491 | * @param string[] $removeTags |
| 492 | * @param string[] $segmentBreakingTags |
| 493 | * @return array |
| 494 | * @throws MWException |
| 495 | */ |
| 496 | protected function cleanHtmlDom( |
| 497 | string $displayTitle, |
| 498 | string $pageContent, |
| 499 | array $removeTags, |
| 500 | array $segmentBreakingTags |
| 501 | ): array { |
| 502 | $cleaner = new Cleaner( $removeTags, $segmentBreakingTags ); |
| 503 | return $cleaner->cleanHtmlDom( $displayTitle, $pageContent ); |
| 504 | } |
| 505 | } |