Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
33.19% |
150 / 452 |
|
20.00% |
4 / 20 |
CRAP | |
0.00% |
0 / 1 |
| TemplateHandler | |
33.19% |
150 / 452 |
|
20.00% |
4 / 20 |
6070.85 | |
0.00% |
0 / 1 |
| __construct | |
100.00% |
12 / 12 |
|
100.00% |
1 / 1 |
1 | |||
| parserFunctionsWrapper | |
0.00% |
0 / 13 |
|
0.00% |
0 / 1 |
20 | |||
| stripIncludeTokens | |
0.00% |
0 / 17 |
|
0.00% |
0 / 1 |
156 | |||
| processToString | |
0.00% |
0 / 36 |
|
0.00% |
0 / 1 |
420 | |||
| isSafeSubst | |
0.00% |
0 / 3 |
|
0.00% |
0 / 1 |
6 | |||
| resolveTemplateTarget | |
40.70% |
35 / 86 |
|
0.00% |
0 / 1 |
122.94 | |||
| flattenAndAppendToks | |
0.00% |
0 / 16 |
|
0.00% |
0 / 1 |
132 | |||
| convertToString | |
0.00% |
0 / 18 |
|
0.00% |
0 / 1 |
6 | |||
| enforceTemplateConstraints | |
100.00% |
10 / 10 |
|
100.00% |
1 / 1 |
2 | |||
| expandTemplateNatively | |
0.00% |
0 / 40 |
|
0.00% |
0 / 1 |
90 | |||
| processTemplateSource | |
73.53% |
25 / 34 |
|
0.00% |
0 / 1 |
3.17 | |||
| encapTokens | |
100.00% |
4 / 4 |
|
100.00% |
1 / 1 |
1 | |||
| processTemplateTokens | |
62.50% |
10 / 16 |
|
0.00% |
0 / 1 |
15.27 | |||
| fetchTemplateAndTitle | |
0.00% |
0 / 10 |
|
0.00% |
0 / 1 |
20 | |||
| hasTemplateToken | |
40.00% |
2 / 5 |
|
0.00% |
0 / 1 |
7.46 | |||
| processSpecialMagicWord | |
0.00% |
0 / 41 |
|
0.00% |
0 / 1 |
56 | |||
| expandTemplate | |
55.22% |
37 / 67 |
|
0.00% |
0 / 1 |
42.94 | |||
| onTemplate | |
100.00% |
10 / 10 |
|
100.00% |
1 / 1 |
3 | |||
| onTemplateArg | |
0.00% |
0 / 8 |
|
0.00% |
0 / 1 |
12 | |||
| onTag | |
83.33% |
5 / 6 |
|
0.00% |
0 / 1 |
4.07 | |||
| 1 | <?php |
| 2 | declare( strict_types = 1 ); |
| 3 | |
| 4 | namespace Wikimedia\Parsoid\Wt2Html\TT; |
| 5 | |
| 6 | use Wikimedia\Assert\Assert; |
| 7 | use Wikimedia\Assert\UnreachableException; |
| 8 | use Wikimedia\Parsoid\Tokens\CommentTk; |
| 9 | use Wikimedia\Parsoid\Tokens\EndTagTk; |
| 10 | use Wikimedia\Parsoid\Tokens\KV; |
| 11 | use Wikimedia\Parsoid\Tokens\NlTk; |
| 12 | use Wikimedia\Parsoid\Tokens\SelfclosingTagTk; |
| 13 | use Wikimedia\Parsoid\Tokens\SourceRange; |
| 14 | use Wikimedia\Parsoid\Tokens\TagTk; |
| 15 | use Wikimedia\Parsoid\Tokens\Token; |
| 16 | use Wikimedia\Parsoid\Utils\DOMCompat; |
| 17 | use Wikimedia\Parsoid\Utils\DOMUtils; |
| 18 | use Wikimedia\Parsoid\Utils\PHPUtils; |
| 19 | use Wikimedia\Parsoid\Utils\PipelineUtils; |
| 20 | use Wikimedia\Parsoid\Utils\Title; |
| 21 | use Wikimedia\Parsoid\Utils\TitleException; |
| 22 | use Wikimedia\Parsoid\Utils\TokenUtils; |
| 23 | use Wikimedia\Parsoid\Utils\Utils; |
| 24 | use Wikimedia\Parsoid\Wikitext\Wikitext; |
| 25 | use Wikimedia\Parsoid\Wt2Html\Params; |
| 26 | use Wikimedia\Parsoid\Wt2Html\TokenTransformManager; |
| 27 | |
| 28 | /** |
| 29 | * Template and template argument handling. |
| 30 | */ |
| 31 | class TemplateHandler extends TokenHandler { |
| 32 | /** |
| 33 | * @var bool Should we wrap template tokens with template meta tags? |
| 34 | */ |
| 35 | private $wrapTemplates; |
| 36 | |
| 37 | /** |
| 38 | * @var AttributeExpander |
| 39 | * Local copy of the attribute expander to deal with template targets |
| 40 | * that are templated themselves |
| 41 | */ |
| 42 | private $ae; |
| 43 | |
| 44 | /** |
| 45 | * @var ParserFunctions |
| 46 | */ |
| 47 | private $parserFunctions; |
| 48 | |
| 49 | /** |
| 50 | * @var bool |
| 51 | */ |
| 52 | private $atMaxArticleSize; |
| 53 | |
| 54 | /** @var string|null */ |
| 55 | private $safeSubstRegex; |
| 56 | |
| 57 | /** |
| 58 | * @param TokenTransformManager $manager |
| 59 | * @param array $options |
| 60 | * - ?bool inTemplate Is this being invoked while processing a template? |
| 61 | * - ?bool expandTemplates Should we expand templates encountered here? |
| 62 | * - ?string extTag The name of the extension tag, if any, which is being expanded. |
| 63 | */ |
| 64 | public function __construct( TokenTransformManager $manager, array $options ) { |
| 65 | parent::__construct( $manager, $options ); |
| 66 | // Set this here so that it's available in the TokenStreamPatcher, |
| 67 | // which continues to inherit from TemplateHandler. |
| 68 | $this->parserFunctions = new ParserFunctions( $this->env ); |
| 69 | $this->ae = new AttributeExpander( $this->manager, [ |
| 70 | 'expandTemplates' => $this->options['expandTemplates'], |
| 71 | 'inTemplate' => $this->options['inTemplate'], |
| 72 | 'standalone' => true, |
| 73 | ] ); |
| 74 | $this->wrapTemplates = !$options['inTemplate']; |
| 75 | |
| 76 | // In the legacy parser, the call to replaceVariables from internalParse |
| 77 | // returns early if the text is already greater than the $wgMaxArticleSize |
| 78 | // We're going to compare and set a boolean here, then do the "early |
| 79 | // return" below. |
| 80 | $this->atMaxArticleSize = !$this->env->compareWt2HtmlLimit( |
| 81 | 'wikitextSize', |
| 82 | strlen( $this->env->topFrame->getSrcText() ) |
| 83 | ); |
| 84 | } |
| 85 | |
| 86 | /** |
| 87 | * Parser functions also need template wrapping. |
| 88 | * |
| 89 | * @param array $tokens |
| 90 | * @return array |
| 91 | */ |
| 92 | private function parserFunctionsWrapper( array $tokens ): array { |
| 93 | $chunkToks = []; |
| 94 | if ( $tokens ) { |
| 95 | // This is only for the Parsoid native expansion pipeline used in |
| 96 | // parser tests. The "" token sometimes changes foster parenting |
| 97 | // behavior and trips up some tests. |
| 98 | $tokens = array_values( array_filter( $tokens, static function ( $t ) { |
| 99 | return $t !== ''; |
| 100 | } ) ); |
| 101 | |
| 102 | // token chunk should be flattened |
| 103 | $flat = true; |
| 104 | foreach ( $tokens as $t ) { |
| 105 | if ( is_array( $t ) ) { |
| 106 | $flat = false; |
| 107 | break; |
| 108 | } |
| 109 | } |
| 110 | Assert::invariant( $flat, "Expected token chunk to be flattened" ); |
| 111 | |
| 112 | $chunkToks = $this->processTemplateTokens( $tokens ); |
| 113 | } |
| 114 | return $chunkToks; |
| 115 | } |
| 116 | |
| 117 | /** |
| 118 | * Strip include tags, and the contents of includeonly tags as well. |
| 119 | * @param (Token|string)[] $tokens |
| 120 | * @return (Token|string)[] |
| 121 | */ |
| 122 | private function stripIncludeTokens( array $tokens ): array { |
| 123 | $toks = []; |
| 124 | $includeOnly = false; |
| 125 | foreach ( $tokens as $tok ) { |
| 126 | if ( is_string( $tok ) ) { |
| 127 | if ( !$includeOnly ) { |
| 128 | $toks[] = $tok; |
| 129 | } |
| 130 | continue; |
| 131 | } |
| 132 | |
| 133 | switch ( get_class( $tok ) ) { |
| 134 | case TagTk::class: |
| 135 | case EndTagTk::class: |
| 136 | case SelfclosingTagTk::class: |
| 137 | $tokName = $tok->getName(); |
| 138 | if ( $tokName === 'noinclude' || $tokName === 'onlyinclude' ) { |
| 139 | break; |
| 140 | } elseif ( $tokName === 'includeonly' ) { |
| 141 | $includeOnly = $tok instanceof TagTk; |
| 142 | break; |
| 143 | } |
| 144 | // Fall through |
| 145 | default: |
| 146 | if ( !$includeOnly ) { |
| 147 | $toks[] = $tok; |
| 148 | } |
| 149 | } |
| 150 | } |
| 151 | return $toks; |
| 152 | } |
| 153 | |
| 154 | /** |
| 155 | * Take output of tokensToString and further postprocess it. |
| 156 | * - If it can be processed to a string which would be a valid template transclusion target, |
| 157 | * the return value will be [ $the_string_value, null ] |
| 158 | * - If not, the return value will be [ $partial_string, $unprocessed_token_array ] |
| 159 | * The caller can then decide if this would be a valid parser function call |
| 160 | * where the unprocessed token array would be part of the first arg to the parser function. |
| 161 | * Ex: With "{{uc:foo [[foo]] {{1x|foo}} bar}}", we return |
| 162 | * [ "uc:foo ", [ wikilink-token, " ", template-token, " bar" ] ] |
| 163 | * |
| 164 | * @param array $tokens |
| 165 | * @return array first element is always a string |
| 166 | */ |
| 167 | private function processToString( array $tokens ): array { |
| 168 | $maybeTarget = TokenUtils::tokensToString( $tokens, true, [ 'retainNLs' => true ] ); |
| 169 | if ( !is_array( $maybeTarget ) ) { |
| 170 | return [ $maybeTarget, null ]; |
| 171 | } |
| 172 | |
| 173 | $buf = $maybeTarget[0]; // Will always be a string |
| 174 | $tgtTokens = $maybeTarget[1]; |
| 175 | $preNlContent = null; |
| 176 | foreach ( $tgtTokens as $i => $ntt ) { |
| 177 | if ( is_string( $ntt ) ) { |
| 178 | $buf .= $ntt; |
| 179 | if ( $preNlContent !== null && !preg_match( '/^\s*$/D', $buf ) ) { |
| 180 | // intervening newline makes this an invalid template target |
| 181 | return [ $preNlContent, array_merge( [ $buf ], array_slice( $tgtTokens, $i ) ) ]; |
| 182 | } |
| 183 | continue; |
| 184 | } |
| 185 | |
| 186 | switch ( get_class( $ntt ) ) { |
| 187 | case SelfclosingTagTk::class: |
| 188 | // Quotes are valid template targets |
| 189 | if ( $ntt->getName() === 'mw-quote' ) { |
| 190 | $buf .= $ntt->getAttribute( 'value' ); |
| 191 | } elseif ( !TokenUtils::isEmptyLineMetaToken( $ntt ) && |
| 192 | $ntt->getName() !== 'template' && |
| 193 | $ntt->getName() !== 'templatearg' |
| 194 | ) { |
| 195 | // We are okay with empty (comment-only) lines, |
| 196 | // {{..}} and {{{..}}} in template targets. |
| 197 | if ( $preNlContent !== null ) { |
| 198 | return [ $preNlContent, array_merge( [ $buf ], array_slice( $tgtTokens, $i ) ) ]; |
| 199 | } else { |
| 200 | return [ $buf, array_slice( $tgtTokens, $i ) ]; |
| 201 | } |
| 202 | } |
| 203 | break; |
| 204 | |
| 205 | case TagTk::class: |
| 206 | case EndTagTk::class: |
| 207 | if ( $preNlContent !== null ) { |
| 208 | return [ $preNlContent, array_merge( [ $buf ], array_slice( $tgtTokens, $i ) ) ]; |
| 209 | } else { |
| 210 | return [ $buf, array_slice( $tgtTokens, $i ) ]; |
| 211 | } |
| 212 | |
| 213 | case CommentTk::class: |
| 214 | // Ignore comments as well |
| 215 | break; |
| 216 | |
| 217 | case NlTk::class: |
| 218 | // Ignore only the leading or trailing newlines |
| 219 | // (modulo whitespace and comments) |
| 220 | // |
| 221 | // If we only have whitespace in $buf thus far, |
| 222 | // the newline can be ignored. But, if we have |
| 223 | // non-ws content in $buf, everything that follows |
| 224 | // can only be ws. |
| 225 | if ( preg_match( '/^\s*$/D', $buf ) ) { |
| 226 | $buf .= "\n"; |
| 227 | break; |
| 228 | } elseif ( $preNlContent === null ) { |
| 229 | // Buffer accumulated content |
| 230 | $preNlContent = $buf; |
| 231 | $buf = "\n"; |
| 232 | break; |
| 233 | } else { |
| 234 | return [ $preNlContent, array_merge( [ $buf ], array_slice( $tgtTokens, $i ) ) ]; |
| 235 | } |
| 236 | |
| 237 | default: |
| 238 | throw new UnreachableException( 'Unexpected token type: ' . get_class( $ntt ) ); |
| 239 | } |
| 240 | } |
| 241 | |
| 242 | // All good! No newline / only whitespace/comments post newline. |
| 243 | return [ $preNlContent . $buf, null ]; |
| 244 | } |
| 245 | |
| 246 | /** |
| 247 | * Is the prefix "safesubst" |
| 248 | * @param string $prefix |
| 249 | * @return bool |
| 250 | */ |
| 251 | private function isSafeSubst( $prefix ): bool { |
| 252 | if ( $this->safeSubstRegex === null ) { |
| 253 | $this->safeSubstRegex = $this->env->getSiteConfig()->getMagicWordMatcher( 'safesubst' ); |
| 254 | } |
| 255 | return (bool)preg_match( $this->safeSubstRegex, $prefix . ':' ); |
| 256 | } |
| 257 | |
| 258 | /** |
| 259 | * @param TemplateEncapsulator $state |
| 260 | * @param string|Token|array $targetToks |
| 261 | * @param SourceRange $srcOffsets |
| 262 | * @return array|null |
| 263 | */ |
| 264 | private function resolveTemplateTarget( |
| 265 | TemplateEncapsulator $state, $targetToks, $srcOffsets |
| 266 | ): ?array { |
| 267 | $additionalToks = null; |
| 268 | if ( is_string( $targetToks ) ) { |
| 269 | $target = $targetToks; |
| 270 | } else { |
| 271 | $toks = !is_array( $targetToks ) ? [ $targetToks ] : $targetToks; |
| 272 | $toks = $this->processToString( $this->stripIncludeTokens( $toks ) ); |
| 273 | list( $target, $additionalToks ) = $toks; |
| 274 | } |
| 275 | |
| 276 | $target = trim( $target ); |
| 277 | $pieces = explode( ':', $target ); |
| 278 | $untrimmedPrefix = $pieces[0]; |
| 279 | $prefix = trim( $pieces[0] ); |
| 280 | |
| 281 | // Parser function names usually (not always) start with a hash |
| 282 | $hasHash = substr( $target, 0, 1 ) === '#'; |
| 283 | // String found after the colon will be the parser function arg |
| 284 | $haveColon = count( $pieces ) > 1; |
| 285 | |
| 286 | // safesubst found in content should be treated as if no modifier were |
| 287 | // present. See https://en.wikipedia.org/wiki/Help:Substitution#The_safesubst:_modifier |
| 288 | if ( $haveColon && $this->isSafeSubst( $prefix ) ) { |
| 289 | $target = substr( $target, strlen( $untrimmedPrefix ) + 1 ); |
| 290 | array_shift( $pieces ); |
| 291 | $untrimmedPrefix = $pieces[0]; |
| 292 | $prefix = trim( $pieces[0] ); |
| 293 | $haveColon = count( $pieces ) > 1; |
| 294 | } |
| 295 | |
| 296 | $env = $this->env; |
| 297 | $siteConfig = $env->getSiteConfig(); |
| 298 | |
| 299 | // Additional tokens are only justifiable in parser functions scenario |
| 300 | if ( !$haveColon && $additionalToks ) { |
| 301 | return null; |
| 302 | } |
| 303 | |
| 304 | $pfArg = ''; |
| 305 | if ( $haveColon ) { |
| 306 | $pfArg = substr( $target, strlen( $untrimmedPrefix ) + 1 ); |
| 307 | if ( $additionalToks ) { |
| 308 | $pfArg = [ $pfArg ]; |
| 309 | PHPUtils::pushArray( $pfArg, $additionalToks ); |
| 310 | } |
| 311 | } |
| 312 | |
| 313 | // Check if we have a magic-word variable. |
| 314 | $magicWordVar = $siteConfig->getMagicWordForVariable( $prefix ) ?? |
| 315 | $siteConfig->getMagicWordForVariable( mb_strtolower( $prefix ) ); |
| 316 | if ( $magicWordVar ) { |
| 317 | $state->variableName = $magicWordVar; |
| 318 | return [ |
| 319 | 'isVariable' => true, |
| 320 | 'magicWordType' => $magicWordVar === '!' ? '!' : null, |
| 321 | 'name' => $magicWordVar, |
| 322 | // FIXME: Some made up synthetic title |
| 323 | 'title' => $env->makeTitleFromURLDecodedStr( "Special:Variable/$magicWordVar" ), |
| 324 | 'pfArg' => $pfArg, |
| 325 | 'srcOffsets' => new SourceRange( |
| 326 | $srcOffsets->start + strlen( $untrimmedPrefix ) + 1, |
| 327 | $srcOffsets->end ), |
| 328 | ]; |
| 329 | } |
| 330 | |
| 331 | // FIXME: Checks for msgnw, msg, raw are missing at this point |
| 332 | |
| 333 | $canonicalFunctionName = null; |
| 334 | if ( $haveColon ) { |
| 335 | $canonicalFunctionName = $siteConfig->getMagicWordForFunctionHook( $prefix ); |
| 336 | } |
| 337 | if ( $canonicalFunctionName === null && $hasHash ) { |
| 338 | // If the target starts with a '#' it can't possibly be a template |
| 339 | // so this must be a "broken" parser function invocation |
| 340 | $canonicalFunctionName = substr( $prefix, 1 ); |
| 341 | // @todo: Flag this as an author error somehow (T314524) |
| 342 | } |
| 343 | if ( $canonicalFunctionName !== null ) { |
| 344 | $state->parserFunctionName = $canonicalFunctionName; |
| 345 | // XXX this is made up. |
| 346 | $syntheticTitle = $env->makeTitleFromURLDecodedStr( |
| 347 | "Special:ParserFunction/$canonicalFunctionName", |
| 348 | $env->getSiteConfig()->canonicalNamespaceId( 'Special' ), |
| 349 | true // No exceptions |
| 350 | ); |
| 351 | // Note that parserFunctionName/$canonicalFunctionName is not |
| 352 | // necessarily a valid title! Parsing rules are pretty generous |
| 353 | // w/r/t valid parser function names. |
| 354 | if ( $syntheticTitle === null ) { |
| 355 | $syntheticTitle = $env->makeTitleFromText( |
| 356 | 'Special:ParserFunction/unknown' |
| 357 | ); |
| 358 | } |
| 359 | return [ |
| 360 | 'name' => $canonicalFunctionName, |
| 361 | 'pfArg' => $pfArg, |
| 362 | 'srcOffsets' => new SourceRange( |
| 363 | $srcOffsets->start + strlen( $untrimmedPrefix ) + 1, |
| 364 | $srcOffsets->end ), |
| 365 | 'isPF' => true, |
| 366 | // FIXME: Some made up synthetic title |
| 367 | 'title' => $syntheticTitle, |
| 368 | 'magicWordType' => isset( Utils::magicMasqs()[$canonicalFunctionName] ) ? 'MASQ' : null, |
| 369 | 'targetToks' => !is_array( $targetToks ) ? [ $targetToks ] : $targetToks, |
| 370 | ]; |
| 371 | } |
| 372 | |
| 373 | // We've exhausted the parser-function scenarios, and we still have additional tokens. |
| 374 | if ( $additionalToks ) { |
| 375 | return null; |
| 376 | } |
| 377 | |
| 378 | // `resolveTitle()` adds the namespace prefix when it resolves fragments |
| 379 | // and relative titles, and a leading colon should resolve to a template |
| 380 | // from the main namespace, hence we omit a default when making a title |
| 381 | $namespaceId = strspn( $target, ':#/.' ) ? |
| 382 | null : $siteConfig->canonicalNamespaceId( 'template' ); |
| 383 | |
| 384 | // Resolve a possibly relative link and |
| 385 | // normalize the target before template processing. |
| 386 | $title = null; |
| 387 | try { |
| 388 | $title = $env->resolveTitle( $target ); |
| 389 | } catch ( TitleException $e ) { |
| 390 | // Invalid template target! |
| 391 | return null; |
| 392 | } |
| 393 | |
| 394 | // Entities in transclusions aren't decoded in the PHP parser |
| 395 | // So, treat the title as a url-decoded string! |
| 396 | $title = $env->makeTitleFromURLDecodedStr( $title, $namespaceId, true ); |
| 397 | if ( !$title ) { |
| 398 | // Invalid template target! |
| 399 | return null; |
| 400 | } |
| 401 | |
| 402 | // data-mw.target.href should be a url |
| 403 | $state->resolvedTemplateTarget = $env->makeLink( $title ); |
| 404 | |
| 405 | return [ |
| 406 | 'magicWordType' => null, |
| 407 | 'name' => $title->getPrefixedDBKey(), |
| 408 | 'title' => $title, |
| 409 | ]; |
| 410 | } |
| 411 | |
| 412 | /** |
| 413 | * Flatten |
| 414 | * @param (Token|string)[] $tokens |
| 415 | * @param ?string $prefix |
| 416 | * @param Token|string|(Token|string)[] $t |
| 417 | * @return array |
| 418 | */ |
| 419 | private function flattenAndAppendToks( |
| 420 | array $tokens, ?string $prefix, $t |
| 421 | ): array { |
| 422 | if ( is_array( $t ) ) { |
| 423 | $len = count( $t ); |
| 424 | if ( $len > 0 ) { |
| 425 | if ( $prefix !== null && $prefix !== '' ) { |
| 426 | $tokens[] = $prefix; |
| 427 | } |
| 428 | PHPUtils::pushArray( $tokens, $t ); |
| 429 | } |
| 430 | } elseif ( is_string( $t ) ) { |
| 431 | $len = strlen( $t ); |
| 432 | if ( $len > 0 ) { |
| 433 | if ( $prefix !== null && $prefix !== '' ) { |
| 434 | $tokens[] = $prefix; |
| 435 | } |
| 436 | $tokens[] = $t; |
| 437 | } |
| 438 | } else { |
| 439 | if ( $prefix !== null && $prefix !== '' ) { |
| 440 | $tokens[] = $prefix; |
| 441 | } |
| 442 | $tokens[] = $t; |
| 443 | } |
| 444 | |
| 445 | return $tokens; |
| 446 | } |
| 447 | |
| 448 | /** |
| 449 | * By default, don't attempt to expand any templates in the wikitext that will be reprocessed. |
| 450 | * |
| 451 | * @param Token $token |
| 452 | * @param bool $expandTemplates |
| 453 | * @return TemplateExpansionResult |
| 454 | */ |
| 455 | private function convertToString( Token $token, bool $expandTemplates = false ): TemplateExpansionResult { |
| 456 | $frame = $this->manager->getFrame(); |
| 457 | $tsr = $token->dataParsoid->tsr; |
| 458 | $src = substr( $token->dataParsoid->src, 1, -1 ); |
| 459 | $startOffset = $tsr->start + 1; |
| 460 | $srcOffsets = new SourceRange( $startOffset, $startOffset + strlen( $src ) ); |
| 461 | |
| 462 | $toks = PipelineUtils::processContentInPipeline( |
| 463 | $this->env, $frame, $src, [ |
| 464 | 'pipelineType' => 'text/x-mediawiki', |
| 465 | 'pipelineOpts' => [ |
| 466 | 'inTemplate' => $this->options['inTemplate'], |
| 467 | 'expandTemplates' => $expandTemplates && $this->options['expandTemplates'], |
| 468 | ], |
| 469 | 'sol' => false, |
| 470 | 'srcOffsets' => $srcOffsets, |
| 471 | ] |
| 472 | ); |
| 473 | TokenUtils::stripEOFTkfromTokens( $toks ); |
| 474 | return new TemplateExpansionResult( array_merge( [ '{' ], $toks, [ '}' ] ), true ); |
| 475 | } |
| 476 | |
| 477 | /** |
| 478 | * Enforce template loops / loop depth limit constraints and emit |
| 479 | * error message if constraints are violated. |
| 480 | * |
| 481 | * @param mixed $target |
| 482 | * @param Title $title |
| 483 | * @param bool $ignoreLoop |
| 484 | * @return ?array |
| 485 | */ |
| 486 | private function enforceTemplateConstraints( $target, Title $title, bool $ignoreLoop ): ?array { |
| 487 | $error = $this->manager->getFrame()->loopAndDepthCheck( |
| 488 | $title, $this->env->getSiteConfig()->getMaxTemplateDepth(), |
| 489 | $ignoreLoop |
| 490 | ); |
| 491 | |
| 492 | return $error ? [ // Loop detected or depth limit exceeded, abort! |
| 493 | new TagTk( 'span', [ new KV( 'class', 'error' ) ] ), |
| 494 | $error, |
| 495 | new SelfclosingTagTk( 'wikilink', [ new KV( 'href', $target, null, '', '' ) ] ), |
| 496 | new EndTagTk( 'span' ), |
| 497 | ] : null; |
| 498 | } |
| 499 | |
| 500 | /** |
| 501 | * Fetch, tokenize and token-transform a template after all arguments and |
| 502 | * the target were expanded. |
| 503 | * |
| 504 | * @param TemplateEncapsulator $state |
| 505 | * @param array $resolvedTgt |
| 506 | * @param array $attribs |
| 507 | * @return TemplateExpansionResult |
| 508 | */ |
| 509 | private function expandTemplateNatively( |
| 510 | TemplateEncapsulator $state, array $resolvedTgt, array $attribs |
| 511 | ): TemplateExpansionResult { |
| 512 | $env = $this->env; |
| 513 | $encap = $this->options['expandTemplates'] && $this->wrapTemplates; |
| 514 | |
| 515 | // XXX: wrap attribs in object with .dict() and .named() methods, |
| 516 | // and each member (key/value) into object with .tokens(), .dom() and |
| 517 | // .wikitext() methods (subclass of Array) |
| 518 | |
| 519 | $target = $resolvedTgt['name']; |
| 520 | if ( isset( $resolvedTgt['isPF'] ) || isset( $resolvedTgt['isVariable'] ) ) { |
| 521 | // FIXME: HARDCODED to core parser function implementations! |
| 522 | // These should go through function hook registrations in the |
| 523 | // ParserTests mock setup ideally. But, it is complicated because the |
| 524 | // Parsoid core parser function versions have "token" versions |
| 525 | // which are incompatible with implementation in FunctionHookHandler |
| 526 | // and FunctionArgs. So, we continue down this hacky path for now. |
| 527 | if ( $target === '=' ) { |
| 528 | $target = 'equal'; // '=' is not a valid character in function names |
| 529 | } |
| 530 | $target = 'pf_' . $target; |
| 531 | // FIXME: Parsoid may not have implemented the parser function natively |
| 532 | // Emit an error message, but encapsulate it so it roundtrips back. |
| 533 | if ( !is_callable( [ $this->parserFunctions, $target ] ) ) { |
| 534 | // FIXME: Consolidate error response format with enforceTemplateConstraints |
| 535 | $err = 'Parser function implementation for ' . $target . ' missing in Parsoid.'; |
| 536 | return new TemplateExpansionResult( [ $err ], false, $encap ); |
| 537 | } |
| 538 | |
| 539 | $pfAttribs = new Params( $attribs ); |
| 540 | $pfAttribs->args[0] = new KV( |
| 541 | // FIXME: This is bogus, but preserves borked b/c |
| 542 | TokenUtils::tokensToString( $resolvedTgt['pfArg'] ), [], |
| 543 | $resolvedTgt['srcOffsets']->expandTsrK() |
| 544 | ); |
| 545 | $env->log( 'debug', 'entering prefix', $target, $state->token ); |
| 546 | $res = call_user_func( [ $this->parserFunctions, $target ], |
| 547 | $state->token, $this->manager->getFrame(), $pfAttribs ); |
| 548 | if ( $this->wrapTemplates ) { |
| 549 | $res = $this->parserFunctionsWrapper( $res ); |
| 550 | } |
| 551 | return new TemplateExpansionResult( $res, false, $encap ); |
| 552 | } |
| 553 | |
| 554 | // Loop detection needs to be enabled since we're doing our own template expansion |
| 555 | $error = $this->enforceTemplateConstraints( $target, $resolvedTgt['title'], false ); |
| 556 | if ( $error ) { |
| 557 | // FIXME: Should we be encapsulating here? |
| 558 | // Inconsistent with the other place constrainsts are enforced. |
| 559 | return new TemplateExpansionResult( $error, false, $encap ); |
| 560 | } |
| 561 | |
| 562 | // XXX: notes from brion's mediawiki.parser.environment |
| 563 | // resolve template name |
| 564 | // load template w/ canonical name |
| 565 | // load template w/ variant names (language variants) |
| 566 | |
| 567 | // Fetch template source and expand it |
| 568 | $src = $this->fetchTemplateAndTitle( $target, $attribs ); |
| 569 | if ( $src !== null ) { |
| 570 | $toks = $this->processTemplateSource( |
| 571 | $state->token, |
| 572 | [ |
| 573 | 'name' => $target, |
| 574 | 'title' => $resolvedTgt['title'], |
| 575 | 'attribs' => array_slice( $attribs, 1 ), // strip template target |
| 576 | ], |
| 577 | $src |
| 578 | ); |
| 579 | return new TemplateExpansionResult( $toks, true, $encap ); |
| 580 | } else { |
| 581 | // Convert to a wikilink (which will become a redlink after the redlinks pass). |
| 582 | $toks = [ new SelfclosingTagTk( 'wikilink' ) ]; |
| 583 | $hrefSrc = $resolvedTgt['name']; |
| 584 | $toks[0]->attribs[] = new KV( 'href', $hrefSrc, null, null, $hrefSrc ); |
| 585 | return new TemplateExpansionResult( $toks, false, $encap ); |
| 586 | } |
| 587 | } |
| 588 | |
| 589 | /** |
| 590 | * Process a fetched template source to a token stream. |
| 591 | * |
| 592 | * @param Token $token |
| 593 | * @param array $tplArgs |
| 594 | * @param string $src |
| 595 | * @return array |
| 596 | */ |
| 597 | private function processTemplateSource( Token $token, array $tplArgs, string $src ): array { |
| 598 | $env = $this->env; |
| 599 | $frame = $this->manager->getFrame(); |
| 600 | if ( $env->hasDumpFlag( 'tplsrc' ) ) { |
| 601 | $dump = str_repeat( '=', 28 ) . " template source " . |
| 602 | str_repeat( '=', 28 ) . "\n"; |
| 603 | $dump .= 'TEMPLATE:' . $tplArgs['name'] . 'TRANSCLUSION:' . |
| 604 | PHPUtils::jsonEncode( $token->dataParsoid->src ) . "\n"; |
| 605 | $dump .= str_repeat( '-', 80 ) . "\n"; |
| 606 | $dump .= $src . "\n"; |
| 607 | $dump .= str_repeat( '-', 80 ) . "\n"; |
| 608 | $env->writeDump( $dump ); |
| 609 | } |
| 610 | |
| 611 | if ( $src === '' ) { |
| 612 | return []; |
| 613 | } |
| 614 | |
| 615 | $env->log( 'debug', 'TemplateHandler.processTemplateSource', |
| 616 | $tplArgs['name'], $tplArgs['attribs'] ); |
| 617 | |
| 618 | // Get a nested transformation pipeline for the wikitext that takes |
| 619 | // us through stages 1-2, with the appropriate pipeline options set. |
| 620 | // |
| 621 | // Simply returning the tokenized source here (which may be correct |
| 622 | // when using the legacy preprocessor because we don't expect to |
| 623 | // tokenize any templates or include directives so skipping those |
| 624 | // handlers should be ok) won't work since the options for the pipeline |
| 625 | // we're in probably aren't what we want. |
| 626 | $toks = PipelineUtils::processContentInPipeline( |
| 627 | $env, |
| 628 | $frame, |
| 629 | $src, |
| 630 | [ |
| 631 | 'pipelineType' => 'text/x-mediawiki', |
| 632 | 'pipelineOpts' => [ |
| 633 | 'inTemplate' => true, |
| 634 | 'isInclude' => true, |
| 635 | // FIXME: In reality, this is broken for parser tests where |
| 636 | // we expand templates natively. We do want all nested templates |
| 637 | // to be expanded. But, setting this to !usePHPPreProcessor seems |
| 638 | // to break a number of tests. Not pursuing this line of enquiry |
| 639 | // for now since this parserTests vs production distinction will |
| 640 | // disappear with parser integration. We'll just bear the stench |
| 641 | // till that time. |
| 642 | // |
| 643 | // NOTE: No expansion required for nested templates. |
| 644 | 'expandTemplates' => false, |
| 645 | 'extTag' => $this->options['extTag'] ?? null |
| 646 | ], |
| 647 | 'srcText' => $src, |
| 648 | 'srcOffsets' => new SourceRange( 0, strlen( $src ) ), |
| 649 | 'tplArgs' => $tplArgs, |
| 650 | // HEADS UP: You might be wondering why we are forcing "sol" => true without |
| 651 | // using information about whether the transclusion is used in a SOL context. |
| 652 | // |
| 653 | // Ex: "foo {{1x|*bar}}" Here, "*bar" is not in SOL context relative to the |
| 654 | // top-level page and so, should it be actually be parsed as a list item? |
| 655 | // |
| 656 | // So, there is a use-case where one could argue that the sol value here |
| 657 | // should be conditioned on the page-level context where "{{1x|*bar}}" showed |
| 658 | // up. So, in this example "foo {{1x|*bar}}, sol would be false and in this |
| 659 | // example "foo\n{{1x|*bar}}", sol would be true. That is effectively how |
| 660 | // the legacy parser behaves. (Ignore T2529 for the moment.) |
| 661 | // |
| 662 | // But, Parsoid is a different beast. Since the Parsoid/JS days, templates |
| 663 | // have been processed asynchronously. So, {{1x|*bar}} would be expanded and |
| 664 | // tokenized before even its preceding context might have been processed. |
| 665 | // From the start, Parsoid has aimed to decouple the processing of fragment |
| 666 | // generators (be it templates, extensions, or something else) from the |
| 667 | // processing of the page they are embedded in. This has been the |
| 668 | // starting point of many a wikitext 2.0 proposal on mediawiki.org; |
| 669 | // see also [[mw:Parsing/Notes/Wikitext_2.0#Implications_of_this_model]]. |
| 670 | // |
| 671 | // The main performance implication is that you can process a transclusion |
| 672 | // concurrently *and* cache the output of {{1x|*bar}} since its output is |
| 673 | // the same no matter where on the page it appears. Without this decoupled |
| 674 | // model, if you got "{{mystery-template-that-takes-30-secs}}{{1x|*bar}}" |
| 675 | // you have to wait 30 secs before you get to expand {{1x|*bar}} |
| 676 | // because you have to wait and see whether the mystery template will |
| 677 | // leave you in SOL state or non-SOL state. |
| 678 | // |
| 679 | // In a stroke of good luck, wikitext editors seem to have agreed |
| 680 | // that it is better for all templates to be expanded in a |
| 681 | // consistent SOL state and not be dependent on their context; |
| 682 | // turn now to phab task T2529 which (via a fragile hack) tried |
| 683 | // to ensure that every template which started with |
| 684 | // start-of-line-sensitive markup was evaluated in a |
| 685 | // start-of-line context (by hackily inserting a newline). Not |
| 686 | // everyone was satisfied with this hack (see T14974), but it's |
| 687 | // been the way things work for over a decade now (as evidenced |
| 688 | // by T14974 never having been "fixed"). |
| 689 | // |
| 690 | // So, while we've established we would prefer *not* to use page |
| 691 | // context to set the initial SOL value for tokenizing the |
| 692 | // template, what *should* the initial SOL value be? |
| 693 | // |
| 694 | // * Treat every transclusion as a fresh document starting in SOL |
| 695 | // state, ie set "sol" => true always. This is supported by |
| 696 | // most current wiki use, and is the intent behind the original |
| 697 | // T2529 hack (although that hack left a number of edge cases, |
| 698 | // described below). |
| 699 | // |
| 700 | // * Use `"sol" => false` for templates -- this was the solution |
| 701 | // rejected by the original T2529 as being contrary to editor |
| 702 | // expectations. |
| 703 | // |
| 704 | // * In the future, one might allow the template itself to |
| 705 | // specify that its initial SOL state should be, using a |
| 706 | // mechanism similar to what might be necessary for typed |
| 707 | // templates. This could also address T14974. This is not |
| 708 | // excluded by Parsoid at this point; but it would probably be |
| 709 | // signaled by a template "return type" which is *not* DOM |
| 710 | // therefore the template wouldn't get parsed "as wikitext" |
| 711 | // (ie, T14974 wants an "attribute-value" return type which is |
| 712 | // a plain string, and some of the wikitext 2.0 proposals |
| 713 | // anticipate a "attribute name/value" dictionary as a possible |
| 714 | // return type). |
| 715 | // |
| 716 | // In support of using sol=>true as the default initial state, |
| 717 | // let's examine the sol-sensitive wikitext constructs, and |
| 718 | // implicitly the corner cases left open by the T2529 hack. (For |
| 719 | // non-sol-sensitive constructs, the initial SOL state is |
| 720 | // irrelevant.) |
| 721 | // |
| 722 | // - SOL-sensitive contructs include lists, headings, indent-pre, |
| 723 | // and table syntax. |
| 724 | // - Of these, only lists, headings, and table syntax are actually handled in |
| 725 | // the PEG tokenizer and are impacted by SOL state. |
| 726 | // - Indent-Pre has its own handler that operates in a full page token context |
| 727 | // and isn't impacted. |
| 728 | // - T2529 effectively means for *#:; (lists) and {| (table start), newlines |
| 729 | // are added which means no matter what value we set here, they will get |
| 730 | // processed in sol state. |
| 731 | // - This leaves us with headings (=), table heading (!), table row (|), and |
| 732 | // table close (|}) syntax that would be impacted by what we set here. |
| 733 | // - Given that table row/heading/close templates are very very common on wikis |
| 734 | // and used for constructing complex tables, sol => true will let us handle |
| 735 | // those without hacks. We aren't fully off the hook there -- see the code |
| 736 | // in TokenStreamPatcher, AttributeExpander, TableFixups that all exist to |
| 737 | // to work around the fact that decoupled processing isn't the wikitext |
| 738 | // default. But, without sol => true, we'll likely be in deeper trouble. |
| 739 | // - But, this can cause some occasional bad parses where "=|!" aren't meant |
| 740 | // to be processed as a sol-wikitext construct. |
| 741 | // - Note also that the workaround for T14974 (ie, the T2529 hack applying |
| 742 | // where sol=false is actually desired) has traditionally been to add an |
| 743 | // initial <nowiki/> which ensures that the "T2529 characters" are not |
| 744 | // initial. There are a number of alternative mechanisms to accomplish |
| 745 | // this (ie, HTML-encode the first character). |
| 746 | // |
| 747 | // To honor the spirit of T2529 it seems plausible to try to lint |
| 748 | // away the remaining corner cases where T2529 does *not* result |
| 749 | // in start-of-line state for template expansion, and to use the |
| 750 | // various workarounds for compatibility in the meantime. |
| 751 | // |
| 752 | // We should also pick *one* of the workarounds for T14974 |
| 753 | // (probably `<nowiki/>` at the first position in the template), |
| 754 | // support that (until a better mechanism exists), and (if |
| 755 | // possible) lint away any others. |
| 756 | 'sol' => true |
| 757 | ] |
| 758 | ); |
| 759 | |
| 760 | return $this->processTemplateTokens( $toks ); |
| 761 | } |
| 762 | |
| 763 | /** |
| 764 | * Process the main template element, including the arguments. |
| 765 | * |
| 766 | * @param TemplateEncapsulator $state |
| 767 | * @param array $tokens |
| 768 | * @return array |
| 769 | */ |
| 770 | private function encapTokens( TemplateEncapsulator $state, array $tokens ): array { |
| 771 | // Template encapsulation normally wouldn't happen in nested context, |
| 772 | // since they should have already been expanded, and indeed we set |
| 773 | // expandTemplates === false in processTemplateSource. However, |
| 774 | // extension tags from templates can have content that requires wikitext |
| 775 | // parsing and, due to precedence, contain unexpanded templates. |
| 776 | // |
| 777 | // For example, {{1x|hi<ref>{{1x|ho}}</ref>}} |
| 778 | // |
| 779 | // Since extensions can require template expansion unconditionally, we can |
| 780 | // end up here inTemplate, in which case the substrings of env.page.src |
| 781 | // used in getArgInfo are no longer accurate, and so tplarginfo should be |
| 782 | // omitted. Presumably, template wrapping in the dom post processor won't |
| 783 | // be happening anyways, so this is unnecessary work as it is. |
| 784 | Assert::invariant( |
| 785 | $this->wrapTemplates, 'Encapsulating tokens when not wrapping!' |
| 786 | ); |
| 787 | return $state->encapTokens( $tokens ); |
| 788 | } |
| 789 | |
| 790 | /** |
| 791 | * Handle chunk emitted from the input pipeline after feeding it a template. |
| 792 | * |
| 793 | * @param array $chunk |
| 794 | * @return array |
| 795 | */ |
| 796 | private function processTemplateTokens( array $chunk ): array { |
| 797 | TokenUtils::stripEOFTkfromTokens( $chunk ); |
| 798 | |
| 799 | foreach ( $chunk as $i => $t ) { |
| 800 | if ( $t && isset( $t->dataParsoid->tsr ) ) { |
| 801 | unset( $t->dataParsoid->tsr ); |
| 802 | } |
| 803 | if ( $t instanceof SelfclosingTagTk && |
| 804 | strtolower( $t->getName() ) === 'meta' && |
| 805 | TokenUtils::hasTypeOf( $t, 'mw:Placeholder' ) |
| 806 | ) { |
| 807 | // replace with empty string to avoid metas being foster-parented out |
| 808 | $chunk[$i] = ''; |
| 809 | } |
| 810 | } |
| 811 | |
| 812 | // FIXME: What is this stuff here? Why do we care about stripping out comments |
| 813 | // so much that we create a new token array for every expanded template? |
| 814 | // Unlikely to help perf very much. |
| 815 | if ( !$this->options['expandTemplates'] ) { |
| 816 | // Ignore comments in template transclusion mode |
| 817 | $newChunk = []; |
| 818 | for ( $i = 0, $n = count( $chunk ); $i < $n; $i++ ) { |
| 819 | if ( !( $chunk[$i] instanceof CommentTk ) ) { |
| 820 | $newChunk[] = $chunk[$i]; |
| 821 | } |
| 822 | } |
| 823 | $chunk = $newChunk; |
| 824 | } |
| 825 | |
| 826 | $this->env->log( 'debug', 'TemplateHandler.processTemplateTokens', $chunk ); |
| 827 | return $chunk; |
| 828 | } |
| 829 | |
| 830 | /** |
| 831 | * Fetch a template. |
| 832 | * |
| 833 | * @param string $templateName |
| 834 | * @param array $attribs |
| 835 | * @return ?string |
| 836 | */ |
| 837 | private function fetchTemplateAndTitle( string $templateName, array $attribs ): ?string { |
| 838 | $env = $this->env; |
| 839 | if ( isset( $env->pageCache[$templateName] ) ) { |
| 840 | return $env->pageCache[$templateName]; |
| 841 | } |
| 842 | |
| 843 | $start = microtime( true ); |
| 844 | $pageContent = $env->getDataAccess()->fetchTemplateSource( $env->getPageConfig(), $templateName ); |
| 845 | if ( $env->profiling() ) { |
| 846 | $profile = $env->getCurrentProfile(); |
| 847 | $profile->bumpMWTime( "TemplateFetch", 1000 * ( microtime( true ) - $start ), "api" ); |
| 848 | $profile->bumpCount( "TemplateFetch" ); |
| 849 | } |
| 850 | |
| 851 | // FIXME: |
| 852 | // 1. Hard-coded 'main' role |
| 853 | return $pageContent ? $pageContent->getContent( 'main' ) : null; |
| 854 | } |
| 855 | |
| 856 | /** |
| 857 | * @param mixed $tokens |
| 858 | * @return bool |
| 859 | */ |
| 860 | private static function hasTemplateToken( $tokens ): bool { |
| 861 | if ( is_array( $tokens ) ) { |
| 862 | foreach ( $tokens as $t ) { |
| 863 | if ( TokenUtils::isTemplateToken( $t ) ) { |
| 864 | return true; |
| 865 | } |
| 866 | } |
| 867 | } |
| 868 | return false; |
| 869 | } |
| 870 | |
| 871 | /** |
| 872 | * Process the special magic word as specified by $resolvedTgt['magicWordType']. |
| 873 | * ``` |
| 874 | * magicWordType === '!' => {{!}} is the magic word |
| 875 | * magicWordtype === 'MASQ' => DEFAULTSORT, DISPLAYTITLE are the magic words |
| 876 | * (See Util::magicMasqs()) |
| 877 | * ``` |
| 878 | * @param bool $atTopLevel |
| 879 | * @param TemplateEncapsulator $state |
| 880 | * @param array $resolvedTgt |
| 881 | * @return TemplateExpansionResult |
| 882 | */ |
| 883 | public function processSpecialMagicWord( |
| 884 | bool $atTopLevel, TemplateEncapsulator $state, array $resolvedTgt |
| 885 | ): TemplateExpansionResult { |
| 886 | $env = $this->env; |
| 887 | $tplToken = $state->token; |
| 888 | |
| 889 | // Special case for {{!}} magic word. |
| 890 | // |
| 891 | // If we tokenized as a magic word, we meant for it to expand to a |
| 892 | // string. The tokenizer has handling for this syntax in table |
| 893 | // positions. However, proceeding to go through template expansion |
| 894 | // will reparse it as a table cell token. Hence this special case |
| 895 | // handling to avoid that path. |
| 896 | if ( $resolvedTgt['magicWordType'] === '!' || $tplToken->attribs[0]->k === '!' ) { |
| 897 | // If we're not at the top level, return a table cell. This will always |
| 898 | // be the case. Either {{!}} was tokenized as a td, or it was tokenized |
| 899 | // as template but the recursive call to fetch its content returns a |
| 900 | // single | in an ambiguous context which will again be tokenized as td. |
| 901 | // In any case, this should only be relevant for parserTests. |
| 902 | if ( empty( $atTopLevel ) ) { |
| 903 | $toks = [ new TagTk( 'td' ) ]; |
| 904 | } else { |
| 905 | $toks = [ '|' ]; |
| 906 | } |
| 907 | return new TemplateExpansionResult( $toks, false, (bool)$this->wrapTemplates ); |
| 908 | } |
| 909 | |
| 910 | Assert::invariant( |
| 911 | $resolvedTgt['magicWordType'] === 'MASQ', |
| 912 | 'Unexpected magicWordType type: ' . $resolvedTgt['magicWordType'] |
| 913 | ); |
| 914 | |
| 915 | $magicWord = mb_strtolower( $resolvedTgt['name'] ); |
| 916 | $pageProp = 'mw:PageProp/'; |
| 917 | if ( $magicWord === 'defaultsort' ) { |
| 918 | $pageProp .= 'category'; |
| 919 | } |
| 920 | $pageProp .= $magicWord; |
| 921 | |
| 922 | $metaToken = new SelfclosingTagTk( 'meta', |
| 923 | [ new KV( 'property', $pageProp ) ], |
| 924 | $tplToken->dataParsoid->clone() |
| 925 | ); |
| 926 | |
| 927 | if ( isset( $tplToken->dataParsoid->tmp->templatedAttribs ) ) { |
| 928 | // See [[mw:Specs/HTML#Generated_attributes_of_HTML_tags]] |
| 929 | // |
| 930 | // For every attribute that has a templated name and/or value, |
| 931 | // AttributeExpander creates a 2-item array for that attribute. |
| 932 | // [ {txt: '..', html: '..'}, { html: '..'} ] |
| 933 | // 'txt' is the plain-text name/value |
| 934 | // 'html' is the HTML-version of the name/value |
| 935 | // |
| 936 | // Massage the templated magic-word info into a similar format. |
| 937 | // In this case, the attribute name is 'content' (implicit) and |
| 938 | // since it is implicit, the name itself cannot be attribute. |
| 939 | // Hence 'html' property is empty. |
| 940 | // |
| 941 | // The attribute value has been templated and is encoded there. |
| 942 | // |
| 943 | // NOTE: If any part of the 'MAGIC_WORD:value' string is templated, |
| 944 | // we consider the magic word as having expanded attributes, rather |
| 945 | // than only when the 'value' part of it. This is because of the |
| 946 | // limitation of our token representation for templates. This is |
| 947 | // an edge case that it is not worth a refactoring right now to |
| 948 | // handle this properly and choose mw:Transclusion or mw:ExpandedAttrs |
| 949 | // depending on which part is templated. |
| 950 | // |
| 951 | // FIXME: Is there a simpler / better repn. for templated attrs? |
| 952 | $ta = $tplToken->dataParsoid->tmp->templatedAttribs; |
| 953 | $html = $ta[0][0]['html']; |
| 954 | $ta[0] = [ |
| 955 | [ 'txt' => 'content' ], // Magic-word attribute name |
| 956 | // FIXME: the content still contains the parser function prefix |
| 957 | // (eg, the html is 'DISPLAYTITLE:Foo' even though the stripped |
| 958 | // content attribute is 'Foo') |
| 959 | [ 'html' => $html ], // HTML repn. of the attribute value |
| 960 | ]; |
| 961 | $metaToken->addAttribute( 'data-mw', PHPUtils::jsonEncode( [ 'attribs' => $ta ] ) ); |
| 962 | |
| 963 | // Use the textContent of the expanded attribute, similar to how |
| 964 | // Sanitizer::sanitizeTagAttr does it. However, here we have the |
| 965 | // opportunity to strip the parser function prefix. |
| 966 | $dom = DOMUtils::parseHTML( $html ); |
| 967 | $content = DOMCompat::getBody( $dom )->textContent; |
| 968 | $content = preg_replace( '#^\w+:#', '', $content, 1 ); |
| 969 | $metaToken->addAttribute( 'content', $content, $resolvedTgt['srcOffsets']->expandTsrV() ); |
| 970 | |
| 971 | $metaToken->addAttribute( 'about', $env->newAboutId() ); |
| 972 | $metaToken->addSpaceSeparatedAttribute( 'typeof', 'mw:ExpandedAttrs' ); |
| 973 | } else { |
| 974 | // Leading/trailing WS should be stripped |
| 975 | // |
| 976 | // This is bogus, but preserves existing functionality |
| 977 | // Clearly we don't have an adequate representation for existing uses |
| 978 | // of the DISPLAYTITLE: magic word. |
| 979 | // phpcs:ignore Generic.Files.LineLength.TooLong |
| 980 | // Ex: {{DISPLAYTITLE:User:<span style="text-transform: lowercase;">MC</span><span style="font-size: 80%;">10</span>/Welcome}} |
| 981 | $key = trim( TokenUtils::tokensToString( $resolvedTgt['pfArg'] ) ); |
| 982 | |
| 983 | $src = $tplToken->dataParsoid->src ?? ''; |
| 984 | if ( $src ) { |
| 985 | // If the token has original wikitext, shadow the sort-key |
| 986 | $origKey = PHPUtils::stripSuffix( preg_replace( '/[^:]+:?/', '', $src, 1 ), '}}' ); |
| 987 | $metaToken->addNormalizedAttribute( 'content', $key, $origKey ); |
| 988 | } else { |
| 989 | // If not, this token came from an extension/template |
| 990 | // in which case, dont bother with shadowing since the token |
| 991 | // will never be edited directly. |
| 992 | $metaToken->addAttribute( 'content', $key ); |
| 993 | } |
| 994 | } |
| 995 | |
| 996 | return new TemplateExpansionResult( [ $metaToken ] ); |
| 997 | } |
| 998 | |
| 999 | /** |
| 1000 | * @param TemplateEncapsulator $state |
| 1001 | * @return TemplateExpansionResult |
| 1002 | */ |
| 1003 | private function expandTemplate( TemplateEncapsulator $state ): TemplateExpansionResult { |
| 1004 | $env = $this->env; |
| 1005 | $token = $state->token; |
| 1006 | $expandTemplates = $this->options['expandTemplates']; |
| 1007 | |
| 1008 | // Since AttributeExpander runs later in the pipeline than TemplateHandler, |
| 1009 | // if the template name is templated, use our copy of AttributeExpander |
| 1010 | // to process all attributes to tokens, and force reprocessing of this |
| 1011 | // template token since we will then know the actual template target. |
| 1012 | if ( $expandTemplates && self::hasTemplateToken( $token->attribs[0]->k ) ) { |
| 1013 | $ret = $this->ae->processComplexAttributes( $token ); |
| 1014 | $toks = $ret->tokens ?? null; |
| 1015 | Assert::invariant( $toks && count( $toks ) === 1 && $toks[0] === $token, |
| 1016 | "Expected only the input token as the return value." ); |
| 1017 | } |
| 1018 | |
| 1019 | if ( $this->atMaxArticleSize ) { |
| 1020 | // As described above, if we were already greater than $wgMaxArticleSize |
| 1021 | // we're going to return the tokens without expanding them. |
| 1022 | // (This case is where the original article as fetched from the DB |
| 1023 | // or passed to the API exceeded max article size.) |
| 1024 | return $this->convertToString( $token ); |
| 1025 | } |
| 1026 | |
| 1027 | // There's no point in proceeding if we've already hit the maximum inclusion size |
| 1028 | // XXX should this be combined with the previous test? |
| 1029 | if ( !$env->bumpWt2HtmlResourceUse( 'wikitextSize', 0 ) ) { |
| 1030 | // FIXME: The legacy parser would try to make this a link and |
| 1031 | // elsewhere we'd return the $e->getMessage() |
| 1032 | // (This case is where the template post-expansion accumulation is |
| 1033 | // over the maximum wikitext size.) |
| 1034 | // XXX: It could be combined with the previous test, but we might |
| 1035 | // want to use different error messages in the future. |
| 1036 | return $this->convertToString( $token ); |
| 1037 | } |
| 1038 | |
| 1039 | $toks = null; |
| 1040 | $text = $token->dataParsoid->src ?? ''; |
| 1041 | |
| 1042 | $tgt = $this->resolveTemplateTarget( |
| 1043 | $state, $token->attribs[0]->k, $token->attribs[0]->srcOffsets->key |
| 1044 | ); |
| 1045 | |
| 1046 | if ( $expandTemplates && $tgt === null ) { |
| 1047 | // Target contains tags, convert template braces and pipes back into text |
| 1048 | // Re-join attribute tokens with '=' and '|' |
| 1049 | return $this->convertToString( $token, true ); |
| 1050 | } |
| 1051 | |
| 1052 | if ( isset( $tgt['magicWordType'] ) ) { |
| 1053 | return $this->processSpecialMagicWord( $this->atTopLevel, $state, $tgt ); |
| 1054 | } |
| 1055 | |
| 1056 | $frame = $this->manager->getFrame(); |
| 1057 | |
| 1058 | if ( $env->nativeTemplateExpansionEnabled() ) { |
| 1059 | // Expand argument keys |
| 1060 | $atm = new AttributeTransformManager( $frame, |
| 1061 | [ 'expandTemplates' => false, 'inTemplate' => true ] |
| 1062 | ); |
| 1063 | $newAttribs = $atm->process( $token->attribs ); |
| 1064 | $target = $newAttribs[0]->k; |
| 1065 | if ( !$target ) { |
| 1066 | $env->log( 'debug', 'No template target! ', $newAttribs ); |
| 1067 | } |
| 1068 | // Resolve the template target again now that the template token's |
| 1069 | // attributes have been expanded by the AttributeTransformManager |
| 1070 | $resolvedTgt = $this->resolveTemplateTarget( $state, $target, $newAttribs[0]->srcOffsets->key ); |
| 1071 | if ( $resolvedTgt === null ) { |
| 1072 | // Target contains tags, convert template braces and pipes back into text |
| 1073 | // Re-join attribute tokens with '=' and '|' |
| 1074 | return $this->convertToString( $token, true ); |
| 1075 | } else { |
| 1076 | return $this->expandTemplateNatively( $state, $resolvedTgt, $newAttribs ); |
| 1077 | } |
| 1078 | } elseif ( $expandTemplates ) { |
| 1079 | // Use MediaWiki's preprocessor |
| 1080 | // |
| 1081 | // The tokenizer needs to use `text` as the cache key for caching |
| 1082 | // expanded tokens from the expanded transclusion text that we get |
| 1083 | // from the preprocessor, since parameter substitution will already |
| 1084 | // have taken place. |
| 1085 | // |
| 1086 | // It's sufficient to pass `[]` in place of attribs since they |
| 1087 | // won't be used. In `usePHPPreProcessor`, there is no parameter |
| 1088 | // substitution coming from the frame. |
| 1089 | |
| 1090 | /* If $tgt is not null, target will be present. */ |
| 1091 | $templateName = $tgt['name']; |
| 1092 | $templateTitle = $tgt['title']; |
| 1093 | // FIXME: This is a source of a lot of issues since templateargs |
| 1094 | // get looked up from the Frame and yield these tokens which then enter |
| 1095 | // the token stream. See T301948 and others from wmf.22 |
| 1096 | // $attribs = array_slice( $token->attribs, 1 ); // Strip template name |
| 1097 | $attribs = []; |
| 1098 | |
| 1099 | // We still need to check for limit violations because of the |
| 1100 | // higher precedence of extension tags, which can result in nested |
| 1101 | // templates even while using the php preprocessor for expansion. |
| 1102 | $error = $this->enforceTemplateConstraints( $templateName, $templateTitle, true ); |
| 1103 | if ( $error ) { |
| 1104 | // FIXME: Should we be encapsulating here? |
| 1105 | // Inconsistent with the other place constrainsts are enforced. |
| 1106 | return new TemplateExpansionResult( $error ); |
| 1107 | } |
| 1108 | |
| 1109 | // Check if we have an expansion for this template in the cache already |
| 1110 | $cachedTransclusion = $env->transclusionCache[$text] ?? null; |
| 1111 | if ( $cachedTransclusion ) { |
| 1112 | // cache hit: reuse the expansion DOM |
| 1113 | // FIXME(SSS): How does this work again for |
| 1114 | // templates like {{start table}} and {[end table}}?? |
| 1115 | return new TemplateExpansionResult( |
| 1116 | PipelineUtils::encapsulateExpansionHTML( |
| 1117 | $env, $token, $cachedTransclusion, [ 'fromCache' => true ] |
| 1118 | ) |
| 1119 | ); |
| 1120 | } else { |
| 1121 | // Fetch and process the template expansion |
| 1122 | $expansion = Wikitext::preprocess( $env, $text ); |
| 1123 | if ( $expansion['error'] ) { |
| 1124 | return new TemplateExpansionResult( |
| 1125 | [ $expansion['src'] ], false, $this->wrapTemplates |
| 1126 | ); |
| 1127 | } else { |
| 1128 | $tplToks = $this->processTemplateSource( |
| 1129 | $token, |
| 1130 | [ |
| 1131 | 'name' => $templateName, |
| 1132 | 'title' => $templateTitle, |
| 1133 | 'attribs' => $attribs |
| 1134 | ], |
| 1135 | $expansion['src'] |
| 1136 | ); |
| 1137 | return new TemplateExpansionResult( |
| 1138 | $tplToks, true, $this->wrapTemplates |
| 1139 | ); |
| 1140 | } |
| 1141 | } |
| 1142 | } else { |
| 1143 | // We don't perform recursive template expansion- something |
| 1144 | // template-like that the PHP parser did not expand. This is |
| 1145 | // encapsulated already, so just return the plain text. |
| 1146 | Assert::invariant( TokenUtils::isTemplateToken( $token ), "Expected template token." ); |
| 1147 | return $this->convertToString( $token ); |
| 1148 | } |
| 1149 | } |
| 1150 | |
| 1151 | /** |
| 1152 | * Main template token handler. |
| 1153 | * |
| 1154 | * Expands target and arguments (both keys and values) and either directly |
| 1155 | * calls or sets up the callback to expandTemplate, which then fetches and |
| 1156 | * processes the template. |
| 1157 | * |
| 1158 | * @param Token $token |
| 1159 | * @return TokenHandlerResult |
| 1160 | */ |
| 1161 | private function onTemplate( Token $token ): TokenHandlerResult { |
| 1162 | $state = new TemplateEncapsulator( |
| 1163 | $this->env, $this->manager->getFrame(), $token, 'mw:Transclusion' |
| 1164 | ); |
| 1165 | $res = $this->expandTemplate( $state ); |
| 1166 | $toks = $res->tokens; |
| 1167 | if ( $res->encap ) { |
| 1168 | $toks = $this->encapTokens( $state, $toks ); |
| 1169 | } |
| 1170 | if ( $res->shuttle ) { |
| 1171 | // Shuttle tokens to the end of the stage since they've gone through the |
| 1172 | // rest of the handlers in the current pipeline in the pipeline above. |
| 1173 | $toks = $this->manager->shuttleTokensToEndOfStage( $toks ); |
| 1174 | } |
| 1175 | return new TokenHandlerResult( $toks ); |
| 1176 | } |
| 1177 | |
| 1178 | /** |
| 1179 | * Expand template arguments with tokens from the containing frame. |
| 1180 | * @param Token $token |
| 1181 | * @return TokenHandlerResult |
| 1182 | */ |
| 1183 | private function onTemplateArg( Token $token ): TokenHandlerResult { |
| 1184 | $toks = $this->manager->getFrame()->expandTemplateArg( $token ); |
| 1185 | |
| 1186 | if ( $this->wrapTemplates && $this->options['expandTemplates'] ) { |
| 1187 | // This is a bare use of template arg syntax at the top level |
| 1188 | // outside any template use context. Wrap this use with RDF attrs. |
| 1189 | // so that this chunk can be RT-ed en-masse. |
| 1190 | $state = new TemplateEncapsulator( |
| 1191 | $this->env, $this->manager->getFrame(), $token, 'mw:Param' |
| 1192 | ); |
| 1193 | $toks = $this->encapTokens( $state, $toks ); |
| 1194 | } |
| 1195 | |
| 1196 | // Shuttle tokens to the end of the stage since they've gone through the |
| 1197 | // rest of the handlers in the current pipeline in the pipeline above. |
| 1198 | $toks = $this->manager->shuttleTokensToEndOfStage( $toks ); |
| 1199 | |
| 1200 | return new TokenHandlerResult( $toks ); |
| 1201 | } |
| 1202 | |
| 1203 | /** |
| 1204 | * @param Token $token |
| 1205 | * @return TokenHandlerResult|null |
| 1206 | */ |
| 1207 | public function onTag( Token $token ): ?TokenHandlerResult { |
| 1208 | switch ( $token->getName() ) { |
| 1209 | case "template": |
| 1210 | return $this->onTemplate( $token ); |
| 1211 | case "templatearg": |
| 1212 | return $this->onTemplateArg( $token ); |
| 1213 | default: |
| 1214 | return null; |
| 1215 | } |
| 1216 | } |
| 1217 | } |