Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
52.15% |
158 / 303 |
|
10.53% |
2 / 19 |
CRAP | |
0.00% |
0 / 1 |
| Parsoid | |
52.15% |
158 / 303 |
|
10.53% |
2 / 19 |
558.96 | |
0.00% |
0 / 1 |
| __construct | |
0.00% |
0 / 2 |
|
0.00% |
0 / 1 |
2 | |||
| version | |
0.00% |
0 / 4 |
|
0.00% |
0 / 1 |
6 | |||
| defaultHTMLVersion | |
0.00% |
0 / 1 |
|
0.00% |
0 / 1 |
2 | |||
| resolveContentVersion | |
0.00% |
0 / 5 |
|
0.00% |
0 / 1 |
20 | |||
| supportsLanguageConversion | |
0.00% |
0 / 1 |
|
0.00% |
0 / 1 |
2 | |||
| setupCommonOptions | |
0.00% |
0 / 16 |
|
0.00% |
0 / 1 |
72 | |||
| parseWikitext | |
0.00% |
0 / 27 |
|
0.00% |
0 / 1 |
56 | |||
| wikitext2html | |
96.77% |
30 / 31 |
|
0.00% |
0 / 1 |
6 | |||
| recordParseMetrics | |
46.55% |
27 / 58 |
|
0.00% |
0 / 1 |
17.77 | |||
| wikitext2lint | |
100.00% |
4 / 4 |
|
100.00% |
1 / 1 |
2 | |||
| dom2wikitext | |
0.00% |
0 / 17 |
|
0.00% |
0 / 1 |
6 | |||
| recordSerializationMetrics | |
69.23% |
18 / 26 |
|
0.00% |
0 / 1 |
4.47 | |||
| html2wikitext | |
100.00% |
3 / 3 |
|
100.00% |
1 / 1 |
1 | |||
| pb2pb | |
97.30% |
72 / 74 |
|
0.00% |
0 / 1 |
7 | |||
| substTopLevelTemplates | |
0.00% |
0 / 3 |
|
0.00% |
0 / 1 |
2 | |||
| findDowngrade | |
0.00% |
0 / 5 |
|
0.00% |
0 / 1 |
20 | |||
| downgrade | |
0.00% |
0 / 14 |
|
0.00% |
0 / 1 |
30 | |||
| implementsLanguageConversionBcp47 | |
80.00% |
4 / 5 |
|
0.00% |
0 / 1 |
2.03 | |||
| downgrade999to2 | |
0.00% |
0 / 7 |
|
0.00% |
0 / 1 |
2 | |||
| 1 | <?php |
| 2 | declare( strict_types = 1 ); |
| 3 | |
| 4 | namespace Wikimedia\Parsoid; |
| 5 | |
| 6 | use Composer\InstalledVersions; |
| 7 | use Composer\Semver\Comparator; |
| 8 | use Composer\Semver\Semver; |
| 9 | use InvalidArgumentException; |
| 10 | use LogicException; |
| 11 | use Wikimedia\Bcp47Code\Bcp47Code; |
| 12 | use Wikimedia\Parsoid\Config\DataAccess; |
| 13 | use Wikimedia\Parsoid\Config\Env; |
| 14 | use Wikimedia\Parsoid\Config\PageConfig; |
| 15 | use Wikimedia\Parsoid\Config\SiteConfig; |
| 16 | use Wikimedia\Parsoid\Config\StubMetadataCollector; |
| 17 | use Wikimedia\Parsoid\Core\ContentMetadataCollector; |
| 18 | use Wikimedia\Parsoid\Core\PageBundle; |
| 19 | use Wikimedia\Parsoid\Core\ResourceLimitExceededException; |
| 20 | use Wikimedia\Parsoid\Core\SelectiveUpdateData; |
| 21 | use Wikimedia\Parsoid\DOM\Document; |
| 22 | use Wikimedia\Parsoid\Ext\ParsoidExtensionAPI; |
| 23 | use Wikimedia\Parsoid\Language\LanguageConverter; |
| 24 | use Wikimedia\Parsoid\Logger\LintLogger; |
| 25 | use Wikimedia\Parsoid\Utils\ComputeSelectiveStats; |
| 26 | use Wikimedia\Parsoid\Utils\ContentUtils; |
| 27 | use Wikimedia\Parsoid\Utils\DOMCompat; |
| 28 | use Wikimedia\Parsoid\Utils\DOMDataUtils; |
| 29 | use Wikimedia\Parsoid\Utils\DOMUtils; |
| 30 | use Wikimedia\Parsoid\Utils\Timing; |
| 31 | use Wikimedia\Parsoid\Utils\Utils; |
| 32 | use Wikimedia\Parsoid\Wikitext\Wikitext; |
| 33 | use Wikimedia\Parsoid\Wt2Html\DOM\Processors\AddRedLinks; |
| 34 | use Wikimedia\Parsoid\Wt2Html\DOM\Processors\ConvertOffsets; |
| 35 | |
| 36 | class Parsoid { |
| 37 | |
| 38 | /** |
| 39 | * Available HTML content versions. |
| 40 | * @see https://www.mediawiki.org/wiki/Parsoid/API#Content_Negotiation |
| 41 | * @see https://www.mediawiki.org/wiki/Specs/HTML#Versioning |
| 42 | */ |
| 43 | public const AVAILABLE_VERSIONS = [ '2.8.0', '999.0.0' ]; |
| 44 | |
| 45 | private const DOWNGRADES = [ |
| 46 | [ 'from' => '999.0.0', 'to' => '2.0.0', 'func' => 'downgrade999to2' ], |
| 47 | ]; |
| 48 | |
| 49 | /** @var SiteConfig */ |
| 50 | private $siteConfig; |
| 51 | |
| 52 | /** @var DataAccess */ |
| 53 | private $dataAccess; |
| 54 | |
| 55 | public function __construct( |
| 56 | SiteConfig $siteConfig, DataAccess $dataAccess |
| 57 | ) { |
| 58 | $this->siteConfig = $siteConfig; |
| 59 | $this->dataAccess = $dataAccess; |
| 60 | } |
| 61 | |
| 62 | /** |
| 63 | * Returns the currently-installed version of Parsoid. |
| 64 | * @return string |
| 65 | */ |
| 66 | public static function version(): string { |
| 67 | try { |
| 68 | // See https://getcomposer.org/doc/07-runtime.md#knowing-the-version-of-package-x |
| 69 | return InstalledVersions::getVersion( 'wikimedia/parsoid' ) ?? |
| 70 | // From the composer runtime API docs: |
| 71 | // "It is nonetheless a good idea to make sure you |
| 72 | // handle the null return value as gracefully as |
| 73 | // possible for safety." |
| 74 | 'null'; |
| 75 | } catch ( \Throwable $t ) { |
| 76 | // Belt-and-suspenders protection against parts of the composer |
| 77 | // runtime API being absent in production. |
| 78 | return 'error'; |
| 79 | } |
| 80 | } |
| 81 | |
| 82 | /** |
| 83 | * Returns the default HTML content version |
| 84 | * @return string |
| 85 | */ |
| 86 | public static function defaultHTMLVersion(): string { |
| 87 | return self::AVAILABLE_VERSIONS[0]; |
| 88 | } |
| 89 | |
| 90 | /** |
| 91 | * See if any content version Parsoid knows how to produce satisfies the |
| 92 | * the supplied version, when interpreted with semver caret semantics. |
| 93 | * This will allow us to make backwards compatible changes, without the need |
| 94 | * for clients to bump the version in their headers all the time. |
| 95 | * |
| 96 | * @param string $version |
| 97 | * @return string|null |
| 98 | */ |
| 99 | public static function resolveContentVersion( string $version ) { |
| 100 | foreach ( self::AVAILABLE_VERSIONS as $i => $a ) { |
| 101 | if ( Semver::satisfies( $a, "^{$version}" ) && |
| 102 | // The section wrapping in 1.6.x should have induced a major |
| 103 | // version bump, since it requires upgrading clients to |
| 104 | // handle it. We therefore hardcode this in so that we can |
| 105 | // fail hard. |
| 106 | Comparator::greaterThanOrEqualTo( $version, '1.6.0' ) |
| 107 | ) { |
| 108 | return $a; |
| 109 | } |
| 110 | } |
| 111 | return null; |
| 112 | } |
| 113 | |
| 114 | /** |
| 115 | * Determine if language conversion is enabled, aka if the optional |
| 116 | * wikimedia/langconv library is installed. |
| 117 | * @return bool True if the wikimedia/langconv library is available |
| 118 | */ |
| 119 | public static function supportsLanguageConversion(): bool { |
| 120 | return class_exists( '\Wikimedia\LangConv\ReplacementMachine' ); |
| 121 | } |
| 122 | |
| 123 | private function setupCommonOptions( array $options ): array { |
| 124 | $envOptions = []; |
| 125 | if ( isset( $options['offsetType'] ) ) { |
| 126 | $envOptions['offsetType'] = $options['offsetType']; |
| 127 | } |
| 128 | if ( isset( $options['traceFlags'] ) ) { |
| 129 | $envOptions['traceFlags'] = $options['traceFlags']; |
| 130 | } |
| 131 | if ( isset( $options['dumpFlags'] ) ) { |
| 132 | $envOptions['dumpFlags'] = $options['dumpFlags']; |
| 133 | } |
| 134 | if ( isset( $options['debugFlags'] ) ) { |
| 135 | $envOptions['debugFlags'] = $options['debugFlags']; |
| 136 | } |
| 137 | if ( !empty( $options['htmlVariantLanguage'] ) ) { |
| 138 | $envOptions['htmlVariantLanguage'] = $options['htmlVariantLanguage']; |
| 139 | } |
| 140 | if ( !empty( $options['wtVariantLanguage'] ) ) { |
| 141 | $envOptions['wtVariantLanguage'] = $options['wtVariantLanguage']; |
| 142 | } |
| 143 | if ( isset( $options['logLevels'] ) ) { |
| 144 | $envOptions['logLevels'] = $options['logLevels']; |
| 145 | } |
| 146 | return $envOptions; |
| 147 | } |
| 148 | |
| 149 | /** |
| 150 | * Parsing code shared between the next two methods. |
| 151 | * |
| 152 | * @param PageConfig $pageConfig |
| 153 | * @param ContentMetadataCollector $metadata |
| 154 | * @param array $options See wikitext2html. |
| 155 | * @param ?SelectiveUpdateData $selparData See wikitext2html. |
| 156 | * @return array{0:Env,1:Document,2:?string} |
| 157 | */ |
| 158 | private function parseWikitext( |
| 159 | PageConfig $pageConfig, |
| 160 | ContentMetadataCollector $metadata, |
| 161 | array $options = [], |
| 162 | ?SelectiveUpdateData $selparData = null |
| 163 | ): array { |
| 164 | $envOptions = $this->setupCommonOptions( $options ); |
| 165 | if ( isset( $options['outputContentVersion'] ) ) { |
| 166 | $envOptions['outputContentVersion'] = $options['outputContentVersion']; |
| 167 | } |
| 168 | $envOptions['discardDataParsoid'] = !empty( $options['discardDataParsoid'] ); |
| 169 | if ( isset( $options['wrapSections'] ) ) { |
| 170 | $envOptions['wrapSections'] = (bool)$options['wrapSections']; |
| 171 | } |
| 172 | if ( isset( $options['pageBundle'] ) ) { |
| 173 | $envOptions['pageBundle'] = (bool)$options['pageBundle']; |
| 174 | } |
| 175 | if ( isset( $options['logLinterData'] ) ) { |
| 176 | $envOptions['logLinterData'] = (bool)$options['logLinterData']; |
| 177 | } |
| 178 | if ( isset( $options['linterOverrides'] ) ) { |
| 179 | $envOptions['linterOverrides'] = $options['linterOverrides']; |
| 180 | } |
| 181 | $envOptions['skipLanguageConversionPass'] = |
| 182 | $options['skipLanguageConversionPass'] ?? false; |
| 183 | |
| 184 | $env = new Env( |
| 185 | $this->siteConfig, $pageConfig, $this->dataAccess, $metadata, $envOptions |
| 186 | ); |
| 187 | if ( !$env->compareWt2HtmlLimit( |
| 188 | 'wikitextSize', strlen( $env->topFrame->getSrcText() ) |
| 189 | ) ) { |
| 190 | throw new ResourceLimitExceededException( |
| 191 | "wt2html: wikitextSize limit exceeded" |
| 192 | ); |
| 193 | } |
| 194 | $contentmodel = $options['contentmodel'] ?? null; |
| 195 | $handler = $env->getContentHandler( $contentmodel ); |
| 196 | $extApi = new ParsoidExtensionAPI( $env ); |
| 197 | // FIXME: Hardcoded to assume 'mode' is 'template' |
| 198 | return [ $env, $handler->toDOM( $extApi, $selparData ), $contentmodel ]; |
| 199 | } |
| 200 | |
| 201 | /** |
| 202 | * Parse the wikitext supplied in a `PageConfig` to HTML. |
| 203 | * |
| 204 | * @param PageConfig $pageConfig |
| 205 | * @param array $options [ |
| 206 | * 'wrapSections' => (bool) Whether `<section>` wrappers should be added. |
| 207 | * 'pageBundle' => (bool) Sets ids on nodes and stores |
| 208 | * data-* attributes in a JSON blob. |
| 209 | * 'body_only' => (bool|null) Only return the <body> children (T181657) |
| 210 | * 'outputContentVersion' => (string|null) Version of HTML to output. |
| 211 | * `null` returns the default version. |
| 212 | * 'contentmodel' => (string|null) The content model of the input. |
| 213 | * 'discardDataParsoid' => (bool) Drop all data-parsoid annotations. |
| 214 | * 'offsetType' => (string) ucs2, char, byte are valid values |
| 215 | * what kind of source offsets should be emitted? |
| 216 | * 'skipLanguageConversionPass' => (bool) Skip the language variant conversion pass (defaults to false) |
| 217 | * 'htmlVariantLanguage' => (Bcp47Code) If non-null, the language variant used for Parsoid HTML. |
| 218 | * 'wtVariantLanguage' => (Bcp47Code) If non-null, the language variant used for wikitext. |
| 219 | * 'logLinterData' => (bool) Should we log linter data if linting is enabled? |
| 220 | * 'linterOverrides' => (array) Override the site linting configs. |
| 221 | * // Debugging options, not for use in production |
| 222 | * 'traceFlags' => (array) associative array with tracing options |
| 223 | * 'dumpFlags' => (array) associative array with dump options |
| 224 | * 'debugFlags' => (array) associative array with debug options |
| 225 | * 'logLevels' => (string[]) Levels to log |
| 226 | * // Experimental options, not considered stable |
| 227 | * 'sampleStats' => (bool) If true, okay to perform "expensive" |
| 228 | * analysis to generate metrics. |
| 229 | * 'renderReason' => (?string) Passed through from MediaWiki core |
| 230 | * to classify metrics; see |
| 231 | * ParserOptions::getRenderReason() |
| 232 | * 'previousInput' => (?PageConfig) wikitext, revision ID, etc of |
| 233 | * some recent parse of this page. |
| 234 | * Not guaranteed to be usable for selective |
| 235 | * update, and could even be from a "newer" |
| 236 | * revision (if this is a render of an old |
| 237 | * revision). |
| 238 | * 'previousOutput' => (?PageBundle) output of the prior parse of |
| 239 | * 'previousInput' |
| 240 | * ] |
| 241 | * @param ?array &$headers |
| 242 | * @param ?ContentMetadataCollector $metadata Pass in a CMC in order to |
| 243 | * collect and retrieve metadata about the parse. |
| 244 | * @param ?SelectiveUpdateData $selparData |
| 245 | * @return PageBundle|string |
| 246 | */ |
| 247 | public function wikitext2html( |
| 248 | PageConfig $pageConfig, array $options = [], ?array &$headers = null, |
| 249 | ?ContentMetadataCollector $metadata = null, ?SelectiveUpdateData $selparData = null |
| 250 | ) { |
| 251 | if ( $metadata === null ) { |
| 252 | $metadata = new StubMetadataCollector( $this->siteConfig ); |
| 253 | } |
| 254 | |
| 255 | $parseTiming = Timing::start(); |
| 256 | [ $env, $doc, $contentmodel ] = $this->parseWikitext( $pageConfig, $metadata, $options, $selparData ); |
| 257 | $parseTimeMs = $parseTiming->end(); |
| 258 | |
| 259 | // FIXME: Does this belong in parseWikitext so that the other endpoint |
| 260 | // is covered as well? It probably depends on expectations of the |
| 261 | // Rest API. If callers of /page/lint/ assume that will update the |
| 262 | // results on the Special page. |
| 263 | if ( $env->linting() ) { |
| 264 | ( new LintLogger( $env ) )->logLintOutput(); |
| 265 | } |
| 266 | |
| 267 | $headers = DOMUtils::findHttpEquivHeaders( $doc ); |
| 268 | $body_only = !empty( $options['body_only'] ); |
| 269 | $node = $body_only ? DOMCompat::getBody( $doc ) : $doc; |
| 270 | |
| 271 | if ( $env->pageBundle ) { |
| 272 | $out = ContentUtils::extractDpAndSerialize( $node, [ |
| 273 | 'innerXML' => $body_only, |
| 274 | ] ); |
| 275 | } else { |
| 276 | $out = [ |
| 277 | 'html' => ContentUtils::toXML( $node, [ |
| 278 | 'innerXML' => $body_only, |
| 279 | ] ), |
| 280 | ]; |
| 281 | } |
| 282 | |
| 283 | $this->recordParseMetrics( |
| 284 | $env, $parseTimeMs, $out, $headers, $contentmodel, $options |
| 285 | ); |
| 286 | |
| 287 | if ( $env->pageBundle ) { |
| 288 | return new PageBundle( |
| 289 | $out['html'], |
| 290 | $out['pb']->parsoid, $out['pb']->mw ?? null, |
| 291 | $env->getOutputContentVersion(), |
| 292 | $headers, |
| 293 | $contentmodel |
| 294 | ); |
| 295 | } else { |
| 296 | return $out['html']; |
| 297 | } |
| 298 | } |
| 299 | |
| 300 | /** |
| 301 | * |
| 302 | */ |
| 303 | private function recordParseMetrics( |
| 304 | Env $env, float $parseTimeMs, |
| 305 | array $out, ?array $headers, string $contentmodel, |
| 306 | array $options |
| 307 | ) { |
| 308 | $metrics = $this->siteConfig->metrics(); |
| 309 | |
| 310 | $pageConfig = $env->getPageConfig(); |
| 311 | |
| 312 | // This is somewhat suspect because ParsoidHandler::tryToCreatePageConfig |
| 313 | // can set a revision id on a MutableRevisionRecord, but it might be simpler |
| 314 | // to make that go away |
| 315 | if ( $pageConfig->getRevisionId() ) { |
| 316 | $mstr = 'pageWithOldid'; |
| 317 | } else { |
| 318 | $mstr = 'wt'; |
| 319 | } |
| 320 | |
| 321 | $timing = Timing::fakeTiming( $this->siteConfig, $parseTimeMs ); |
| 322 | $timing->end( "entry.wt2html.{$mstr}.parse", 'wt2html_parse_seconds', [ 'type' => $mstr ] ); |
| 323 | $version = 'default'; |
| 324 | |
| 325 | if ( Semver::satisfies( |
| 326 | $env->getOutputContentVersion(), '!=' . self::defaultHTMLVersion() |
| 327 | ) ) { |
| 328 | if ( $metrics ) { |
| 329 | $metrics->increment( 'entry.wt2html.parse.version.notdefault' ); |
| 330 | } |
| 331 | $version = 'non-default'; |
| 332 | } |
| 333 | |
| 334 | $this->siteConfig->incrementCounter( 'wt2html_parse_total', [ |
| 335 | 'type' => $mstr, |
| 336 | 'version' => $version |
| 337 | ] ); |
| 338 | |
| 339 | // @phan-suppress-next-line PhanDeprecatedFunction |
| 340 | $timing = Timing::fakeTiming( $this->siteConfig, strlen( $pageConfig->getPageMainContent() ) ); |
| 341 | $timing->end( |
| 342 | "entry.wt2html.{$mstr}.size.input", |
| 343 | "wt2html_size_input_bytes", |
| 344 | [ "type" => $mstr ] |
| 345 | ); |
| 346 | |
| 347 | $outSize = strlen( $out['html'] ); |
| 348 | $timing = Timing::fakeTiming( $this->siteConfig, $outSize ); |
| 349 | $timing->end( "entry.wt2html.{$mstr}.size.output", "wt2html_size_output_bytes", [ "type" => $mstr ] ); |
| 350 | |
| 351 | if ( $parseTimeMs > 10 && $outSize > 100 ) { |
| 352 | // * Don't bother with this metric for really small parse times |
| 353 | // p99 for initialization time is ~7ms according to grafana. |
| 354 | // So, 10ms ensures that startup overheads don't skew the metrics |
| 355 | // * For body_only=false requests, <head> section isn't generated |
| 356 | // and if the output is small, per-request overheads can skew |
| 357 | // the timePerKB metrics. |
| 358 | // |
| 359 | // NOTE: This is slightly misleading since there are fixed costs |
| 360 | // for generating output like the <head> section and should be factored in, |
| 361 | // but this is good enough for now as a useful first degree of approxmation. |
| 362 | $msPerKB = $parseTimeMs * 1024 / $outSize; |
| 363 | $timing = Timing::fakeTiming( $this->siteConfig, $msPerKB ); |
| 364 | $timing->end( |
| 365 | 'entry.wt2html.timePerKB', |
| 366 | 'wt2html_msPerKB', |
| 367 | [] |
| 368 | ); |
| 369 | } |
| 370 | |
| 371 | // Expensive analyses: sampleStats is randomly sampled will not be |
| 372 | // true "often" |
| 373 | $doSample = $options['sampleStats'] ?? false; |
| 374 | if ( !$doSample ) { |
| 375 | return; |
| 376 | } |
| 377 | |
| 378 | try { |
| 379 | // create new page bundle for this computation to ensure we |
| 380 | // don't inadvertently corrupt the main document result. |
| 381 | $newPb = new PageBundle( |
| 382 | $out['html'], |
| 383 | $out['pb']->parsoid, $out['pb']->mw ?? null, |
| 384 | $env->getOutputContentVersion(), |
| 385 | $headers, |
| 386 | $contentmodel |
| 387 | ); |
| 388 | $labels = ComputeSelectiveStats::classify( |
| 389 | $env, |
| 390 | $options['previousInput'] ?? null, |
| 391 | $options['previousOutput'] ?? null, |
| 392 | $pageConfig, |
| 393 | $newPb |
| 394 | ); |
| 395 | $labels['wiki'] = $this->siteConfig->iwp(); |
| 396 | $labels['reason'] = $options['renderReason'] ?? 'unknown'; |
| 397 | |
| 398 | $this->siteConfig->incrementCounter( 'selective_update_total', $labels ); |
| 399 | $this->siteConfig->incrementCounter( 'selective_update_seconds', $labels, $parseTimeMs / 1000. ); |
| 400 | } catch ( \Throwable $t ) { |
| 401 | // Don't ever allow bugs in the classification code to |
| 402 | // impact the availability of content for read views/editing, |
| 403 | // just log. |
| 404 | $env->log( 'warn', 'Classification failure', $t->getTraceAsString() ); |
| 405 | } |
| 406 | } |
| 407 | |
| 408 | /** |
| 409 | * Lint the wikitext supplied in a `PageConfig`. |
| 410 | * |
| 411 | * @param PageConfig $pageConfig |
| 412 | * @param array $options See wikitext2html. |
| 413 | * @param ?ContentMetadataCollector $metadata Pass in a CMC in order to |
| 414 | * collect and retrieve metadata about the parse. |
| 415 | * @return array |
| 416 | */ |
| 417 | public function wikitext2lint( |
| 418 | PageConfig $pageConfig, array $options = [], |
| 419 | ?ContentMetadataCollector $metadata = null |
| 420 | ): array { |
| 421 | if ( $metadata === null ) { |
| 422 | $metadata = new StubMetadataCollector( $this->siteConfig ); |
| 423 | } |
| 424 | [ $env, ] = $this->parseWikitext( $pageConfig, $metadata, $options ); |
| 425 | return $env->getLints(); |
| 426 | } |
| 427 | |
| 428 | /** |
| 429 | * Serialize DOM to wikitext. |
| 430 | * |
| 431 | * @param PageConfig $pageConfig |
| 432 | * @param Document $doc Data attributes are expected to have been applied |
| 433 | * already. Loading them will happen once the environment is created. |
| 434 | * @param array $options [ |
| 435 | * 'inputContentVersion' => (string) The content version of the input. |
| 436 | * Necessary if it differs from the current default in order to |
| 437 | * account for any serialization differences. |
| 438 | * 'offsetType' => (string) ucs2, char, byte are valid values |
| 439 | * what kind of source offsets are present in the HTML? |
| 440 | * 'contentmodel' => (string|null) The content model of the input. |
| 441 | * 'htmlVariantLanguage' => (Bcp47Code) If non-null, the language variant used for Parsoid HTML. |
| 442 | * 'wtVariantLanguage' => (Bcp47Code) If non-null, the language variant used for wikitext. |
| 443 | * 'traceFlags' => (array) associative array with tracing options |
| 444 | * 'dumpFlags' => (array) associative array with dump options |
| 445 | * 'debugFlags' => (array) associative array with debug options |
| 446 | * 'logLevels' => (string[]) Levels to log |
| 447 | * 'htmlSize' => (int) Size of the HTML that generated $doc |
| 448 | * ] |
| 449 | * @param ?SelectiveUpdateData $selserData |
| 450 | * @return string |
| 451 | */ |
| 452 | public function dom2wikitext( |
| 453 | PageConfig $pageConfig, Document $doc, array $options = [], |
| 454 | ?SelectiveUpdateData $selserData = null |
| 455 | ): string { |
| 456 | $envOptions = $this->setupCommonOptions( $options ); |
| 457 | if ( isset( $options['inputContentVersion'] ) ) { |
| 458 | $envOptions['inputContentVersion'] = $options['inputContentVersion']; |
| 459 | } |
| 460 | $envOptions['topLevelDoc'] = $doc; |
| 461 | $metadata = new StubMetadataCollector( $this->siteConfig ); |
| 462 | $env = new Env( |
| 463 | $this->siteConfig, $pageConfig, $this->dataAccess, $metadata, $envOptions |
| 464 | ); |
| 465 | $env->bumpHtml2WtResourceUse( 'htmlSize', $options['htmlSize'] ?? 0 ); |
| 466 | $contentmodel = $options['contentmodel'] ?? null; |
| 467 | $handler = $env->getContentHandler( $contentmodel ); |
| 468 | $extApi = new ParsoidExtensionAPI( $env ); |
| 469 | |
| 470 | $serialTiming = Timing::start(); |
| 471 | $wikitext = $handler->fromDOM( $extApi, $selserData ); |
| 472 | $serialTime = $serialTiming->end(); |
| 473 | |
| 474 | $this->recordSerializationMetrics( $options, $serialTime, $wikitext ); |
| 475 | |
| 476 | return $wikitext; |
| 477 | } |
| 478 | |
| 479 | /** |
| 480 | * |
| 481 | */ |
| 482 | private function recordSerializationMetrics( |
| 483 | array $options, float $serialTime, string $wikitext |
| 484 | ) { |
| 485 | $siteConfig = $this->siteConfig; |
| 486 | $metrics = $siteConfig->metrics(); |
| 487 | |
| 488 | $htmlSize = $options['htmlSize'] ?? 0; |
| 489 | $timing = Timing::fakeTiming( $this->siteConfig, $htmlSize ); |
| 490 | $timing->end( 'entry.html2wt.size.input', 'html2wt_size_input_bytes' ); |
| 491 | |
| 492 | if ( isset( $options['inputContentVersion'] ) ) { |
| 493 | if ( $metrics ) { |
| 494 | $metrics->increment( |
| 495 | 'entry.html2wt.original.version.' . $options['inputContentVersion'] |
| 496 | ); |
| 497 | } |
| 498 | $this->siteConfig->incrementCounter( |
| 499 | 'html2wt_original_version', |
| 500 | [ 'input_content_version' => $options['inputContentVersion'] ] |
| 501 | ); |
| 502 | } |
| 503 | |
| 504 | $timing = Timing::fakeTiming( $this->siteConfig, $serialTime ); |
| 505 | $timing->end( 'entry.html2wt.total', 'html2wt_total_seconds', [] ); |
| 506 | |
| 507 | $timing = Timing::fakeTiming( $this->siteConfig, strlen( $wikitext ) ); |
| 508 | $timing->end( 'entry.html2wt.size.output', 'html2wt_size_output_bytes', [] ); |
| 509 | |
| 510 | if ( $htmlSize ) { // Avoid division by zero |
| 511 | // NOTE: the name timePerInputKB is misleading, since $htmlSize is |
| 512 | // in characters, not bytes. |
| 513 | $msPerKB = $serialTime * 1024 / $htmlSize; |
| 514 | $timing = Timing::fakeTiming( $this->siteConfig, $msPerKB ); |
| 515 | $timing->end( |
| 516 | 'entry.html2wt.timePerInputKB', |
| 517 | 'html2wt_msPerKB', |
| 518 | [] |
| 519 | ); |
| 520 | } |
| 521 | } |
| 522 | |
| 523 | /** |
| 524 | * Serialize HTML to wikitext. Convenience method for dom2wikitext. |
| 525 | * |
| 526 | * @param PageConfig $pageConfig |
| 527 | * @param string $html |
| 528 | * @param array $options |
| 529 | * @param ?SelectiveUpdateData $selserData |
| 530 | * @return string |
| 531 | */ |
| 532 | public function html2wikitext( |
| 533 | PageConfig $pageConfig, string $html, array $options = [], |
| 534 | ?SelectiveUpdateData $selserData = null |
| 535 | ): string { |
| 536 | $doc = DOMUtils::parseHTML( $html, true ); |
| 537 | $options['htmlSize'] ??= mb_strlen( $html ); |
| 538 | return $this->dom2wikitext( $pageConfig, $doc, $options, $selserData ); |
| 539 | } |
| 540 | |
| 541 | /** |
| 542 | * Update the supplied PageBundle based on the `$update` type. |
| 543 | * |
| 544 | * 'convertoffsets': Convert offsets between formats (byte, char, ucs2) |
| 545 | * 'redlinks': Refreshes the classes of known, missing, etc. links. |
| 546 | * 'variant': Converts the HTML based on the supplied variant. |
| 547 | * |
| 548 | * Note that these are DOM transforms, and not roundtrips through wikitext. |
| 549 | * |
| 550 | * @param PageConfig $pageConfig |
| 551 | * @param string $update 'redlinks'|'variant' |
| 552 | * @param PageBundle $pb |
| 553 | * @param array $options |
| 554 | * @return PageBundle |
| 555 | */ |
| 556 | public function pb2pb( |
| 557 | PageConfig $pageConfig, string $update, PageBundle $pb, |
| 558 | array $options = [] |
| 559 | ): PageBundle { |
| 560 | $envOptions = [ |
| 561 | 'pageBundle' => true, |
| 562 | 'topLevelDoc' => DOMUtils::parseHTML( $pb->toHtml(), true ), |
| 563 | ]; |
| 564 | $metadata = new StubMetadataCollector( $this->siteConfig ); |
| 565 | $env = new Env( |
| 566 | $this->siteConfig, $pageConfig, $this->dataAccess, $metadata, $envOptions |
| 567 | ); |
| 568 | $doc = $env->topLevelDoc; |
| 569 | DOMDataUtils::visitAndLoadDataAttribs( |
| 570 | DOMCompat::getBody( $doc ), [ 'markNew' => true ] |
| 571 | ); |
| 572 | |
| 573 | $dataBagPB = DOMDataUtils::getPageBundle( $doc ); |
| 574 | switch ( $update ) { |
| 575 | case 'convertoffsets': |
| 576 | ContentUtils::convertOffsets( |
| 577 | $env, $doc, $options['inputOffsetType'], $options['outputOffsetType'] |
| 578 | ); |
| 579 | $dataBagPB->parsoid['offsetType'] = $options['outputOffsetType']; |
| 580 | $dataBagPB->parsoid['counter'] = $pb->parsoid['counter']; |
| 581 | break; |
| 582 | |
| 583 | case 'redlinks': |
| 584 | ContentUtils::convertOffsets( |
| 585 | $env, $doc, $env->getRequestOffsetType(), 'byte' |
| 586 | ); |
| 587 | ( new AddRedLinks() )->run( $env, DOMCompat::getBody( $doc ) ); |
| 588 | ( new ConvertOffsets() )->run( $env, DOMCompat::getBody( $doc ), [], true ); |
| 589 | break; |
| 590 | |
| 591 | case 'variant': |
| 592 | ContentUtils::convertOffsets( |
| 593 | $env, $doc, $env->getRequestOffsetType(), 'byte' |
| 594 | ); |
| 595 | |
| 596 | // Note that `maybeConvert` could still be a no-op, in case the |
| 597 | // __NOCONTENTCONVERT__ magic word is present, or the htmlVariant |
| 598 | // is a base language code or otherwise invalid. |
| 599 | $hasWtVariant = $options['variant']['wikitext'] ?? |
| 600 | // Deprecated name for this option: |
| 601 | $options['variant']['source'] ?? false; |
| 602 | LanguageConverter::maybeConvert( |
| 603 | $env, $doc, |
| 604 | Utils::mwCodeToBcp47( |
| 605 | $options['variant']['html'] ?? |
| 606 | // Deprecated name for this option: |
| 607 | $options['variant']['target'], |
| 608 | // Be strict in what we accept. |
| 609 | true, $this->siteConfig->getLogger() |
| 610 | ), |
| 611 | $hasWtVariant ? |
| 612 | Utils::mwCodeToBcp47( |
| 613 | $options['variant']['wikitext'] ?? |
| 614 | // Deprecated name for this option: |
| 615 | $options['variant']['source'], |
| 616 | // Be strict in what we accept. |
| 617 | true, $this->siteConfig->getLogger() |
| 618 | ) : null |
| 619 | ); |
| 620 | |
| 621 | // NOTE: Keep this in sync with code in core's LanguageVariantConverter |
| 622 | // Update content-language and vary headers. |
| 623 | DOMUtils::addHttpEquivHeaders( $doc, [ |
| 624 | 'content-language' => $env->htmlContentLanguageBcp47()->toBcp47Code(), |
| 625 | 'vary' => $env->htmlVary() |
| 626 | ] ); |
| 627 | |
| 628 | ( new ConvertOffsets() )->run( $env, DOMCompat::getBody( $doc ), [], true ); |
| 629 | break; |
| 630 | |
| 631 | default: |
| 632 | throw new LogicException( $update . 'is an unknown transformation' ); |
| 633 | } |
| 634 | |
| 635 | DOMDataUtils::visitAndStoreDataAttribs( |
| 636 | DOMCompat::getBody( $doc ), [ |
| 637 | 'discardDataParsoid' => $env->discardDataParsoid, |
| 638 | 'storeInPageBundle' => $env->pageBundle, |
| 639 | 'env' => $env, |
| 640 | ] |
| 641 | ); |
| 642 | $body_only = !empty( $options['body_only'] ); |
| 643 | $node = $body_only ? DOMCompat::getBody( $doc ) : $doc; |
| 644 | DOMDataUtils::injectPageBundle( $doc, $dataBagPB ); |
| 645 | $out = ContentUtils::extractDpAndSerialize( $node, [ |
| 646 | 'innerXML' => $body_only, |
| 647 | ] ); |
| 648 | return new PageBundle( |
| 649 | $out['html'], |
| 650 | $out['pb']->parsoid, $out['pb']->mw ?? null, |
| 651 | // Prefer the passed in version, since this was just a transformation |
| 652 | $pb->version ?? $env->getOutputContentVersion(), |
| 653 | DOMUtils::findHttpEquivHeaders( $doc ), |
| 654 | // Prefer the passed in content model |
| 655 | $pb->contentmodel ?? $pageConfig->getContentModel() |
| 656 | ); |
| 657 | } |
| 658 | |
| 659 | /** |
| 660 | * Perform pre-save transformations with top-level templates subst'd. |
| 661 | * |
| 662 | * @param PageConfig $pageConfig |
| 663 | * @param string $wikitext |
| 664 | * @return string |
| 665 | */ |
| 666 | public function substTopLevelTemplates( |
| 667 | PageConfig $pageConfig, string $wikitext |
| 668 | ): string { |
| 669 | $metadata = new StubMetadataCollector( $this->siteConfig ); |
| 670 | $env = new Env( $this->siteConfig, $pageConfig, $this->dataAccess, $metadata ); |
| 671 | return Wikitext::pst( $env, $wikitext, true /* $substTLTemplates */ ); |
| 672 | } |
| 673 | |
| 674 | /** |
| 675 | * Check whether a given content version can be downgraded to the requested |
| 676 | * content version. |
| 677 | * |
| 678 | * @param string $from Current content version |
| 679 | * @param string $to Requested content version |
| 680 | * @return string[]|null The downgrade that will fulfill the request, as |
| 681 | * [ 'from' => <old version>, 'to' => <new version> ], or null if it |
| 682 | * can't be fulfilled. |
| 683 | */ |
| 684 | public static function findDowngrade( string $from, string $to ): ?array { |
| 685 | foreach ( self::DOWNGRADES as [ 'from' => $dgFrom, 'to' => $dgTo ] ) { |
| 686 | if ( |
| 687 | Semver::satisfies( $from, "^$dgFrom" ) && |
| 688 | Semver::satisfies( $to, "^$dgTo" ) |
| 689 | ) { |
| 690 | // FIXME: Make this a class? |
| 691 | return [ 'from' => $dgFrom, 'to' => $dgTo ]; |
| 692 | } |
| 693 | } |
| 694 | return null; |
| 695 | } |
| 696 | |
| 697 | /** |
| 698 | * Downgrade a document to an older content version. |
| 699 | * |
| 700 | * @param string[] $dg Value returned by findDowngrade(). |
| 701 | * @param PageBundle $pageBundle |
| 702 | */ |
| 703 | public static function downgrade( |
| 704 | array $dg, PageBundle $pageBundle |
| 705 | ): void { |
| 706 | foreach ( self::DOWNGRADES as [ 'from' => $dgFrom, 'to' => $dgTo, 'func' => $dgFunc ] ) { |
| 707 | if ( $dg['from'] === $dgFrom && $dg['to'] === $dgTo ) { |
| 708 | call_user_func( [ self::class, $dgFunc ], $pageBundle ); |
| 709 | |
| 710 | // FIXME: Maybe this resolve should just be part of the $dg |
| 711 | $pageBundle->version = self::resolveContentVersion( $dg['to'] ); |
| 712 | |
| 713 | // FIXME: Maybe this should be a helper to avoid the rt |
| 714 | $doc = DOMUtils::parseHTML( $pageBundle->html ); |
| 715 | // Match the http-equiv meta to the content-type header |
| 716 | $meta = DOMCompat::querySelector( $doc, |
| 717 | 'meta[property="mw:htmlVersion"], meta[property="mw:html:version"]' ); |
| 718 | if ( $meta ) { |
| 719 | $meta->setAttribute( 'content', $pageBundle->version ); |
| 720 | $pageBundle->html = ContentUtils::toXML( $doc ); |
| 721 | } |
| 722 | |
| 723 | return; |
| 724 | } |
| 725 | } |
| 726 | throw new InvalidArgumentException( |
| 727 | "Unsupported downgrade: {$dg['from']} -> {$dg['to']}" |
| 728 | ); |
| 729 | } |
| 730 | |
| 731 | /** |
| 732 | * Check if language variant conversion is implemented for a language |
| 733 | * |
| 734 | * @internal FIXME: Remove once Parsoid's language variant work is completed |
| 735 | * @param PageConfig $pageConfig |
| 736 | * @param Bcp47Code $htmlVariant Variant language to check |
| 737 | * @return bool |
| 738 | */ |
| 739 | public function implementsLanguageConversionBcp47( PageConfig $pageConfig, Bcp47Code $htmlVariant ): bool { |
| 740 | // Hardcode disable zh lang conversion support since Parsoid's |
| 741 | // implementation is incomplete and not performant (T346657). |
| 742 | if ( $pageConfig->getPageLanguageBcp47()->toBcp47Code() === 'zh' ) { |
| 743 | return false; |
| 744 | } |
| 745 | |
| 746 | $metadata = new StubMetadataCollector( $this->siteConfig ); |
| 747 | $env = new Env( $this->siteConfig, $pageConfig, $this->dataAccess, $metadata ); |
| 748 | return LanguageConverter::implementsLanguageConversionBcp47( $env, $htmlVariant ); |
| 749 | } |
| 750 | |
| 751 | /** |
| 752 | * Downgrade the given document and pagebundle from 999.x to 2.x. |
| 753 | * |
| 754 | * @param PageBundle $pageBundle |
| 755 | */ |
| 756 | private static function downgrade999to2( PageBundle $pageBundle ) { |
| 757 | // Effectively, skip applying data-parsoid. Note that if we were to |
| 758 | // support a pb2html downgrade, we'd need to apply the full thing, |
| 759 | // but that would create complications where ids would be left behind. |
| 760 | // See the comment in around `DOMDataUtils::applyPageBundle` |
| 761 | $newPageBundle = new PageBundle( |
| 762 | $pageBundle->html, |
| 763 | [ 'ids' => [] ], |
| 764 | $pageBundle->mw |
| 765 | ); |
| 766 | $pageBundle->html = $newPageBundle->toHtml(); |
| 767 | // Now, modify the pagebundle to the expected form. This is important |
| 768 | // since, at least in the serialization path, the original pb will be |
| 769 | // applied to the modified content and its presence could cause lost |
| 770 | // deletions. |
| 771 | $pageBundle->mw = [ 'ids' => [] ]; |
| 772 | } |
| 773 | } |