Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
0.00% |
0 / 275 |
|
0.00% |
0 / 10 |
CRAP | |
0.00% |
0 / 1 |
| AttributeExpander | |
0.00% |
0 / 275 |
|
0.00% |
0 / 10 |
10506 | |
0.00% |
0 / 1 |
| __construct | |
0.00% |
0 / 2 |
|
0.00% |
0 / 1 |
2 | |||
| nlTkIndex | |
0.00% |
0 / 13 |
|
0.00% |
0 / 1 |
72 | |||
| splitTokens | |
0.00% |
0 / 35 |
|
0.00% |
0 / 1 |
132 | |||
| stripMetaTags | |
0.00% |
0 / 34 |
|
0.00% |
0 / 1 |
210 | |||
| tplToksToString | |
0.00% |
0 / 8 |
|
0.00% |
0 / 1 |
20 | |||
| buildExpandedAttrs | |
0.00% |
0 / 139 |
|
0.00% |
0 / 1 |
2352 | |||
| processComplexAttributes | |
0.00% |
0 / 13 |
|
0.00% |
0 / 1 |
30 | |||
| expandFirstAttribute | |
0.00% |
0 / 14 |
|
0.00% |
0 / 1 |
6 | |||
| onCompoundTk | |
0.00% |
0 / 3 |
|
0.00% |
0 / 1 |
6 | |||
| onAny | |
0.00% |
0 / 14 |
|
0.00% |
0 / 1 |
56 | |||
| 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\Config\Env; |
| 9 | use Wikimedia\Parsoid\NodeData\DataMw; |
| 10 | use Wikimedia\Parsoid\NodeData\DataMwAttrib; |
| 11 | use Wikimedia\Parsoid\Tokens\CompoundTk; |
| 12 | use Wikimedia\Parsoid\Tokens\EmptyLineTk; |
| 13 | use Wikimedia\Parsoid\Tokens\KV; |
| 14 | use Wikimedia\Parsoid\Tokens\NlTk; |
| 15 | use Wikimedia\Parsoid\Tokens\SelfclosingTagTk; |
| 16 | use Wikimedia\Parsoid\Tokens\TagTk; |
| 17 | use Wikimedia\Parsoid\Tokens\Token; |
| 18 | use Wikimedia\Parsoid\Tokens\XMLTagTk; |
| 19 | use Wikimedia\Parsoid\Utils\PHPUtils; |
| 20 | use Wikimedia\Parsoid\Utils\PipelineUtils; |
| 21 | use Wikimedia\Parsoid\Utils\TokenUtils; |
| 22 | use Wikimedia\Parsoid\Utils\Utils; |
| 23 | use Wikimedia\Parsoid\Utils\WTUtils; |
| 24 | use Wikimedia\Parsoid\Wt2Html\Frame; |
| 25 | use Wikimedia\Parsoid\Wt2Html\PegTokenizer; |
| 26 | use Wikimedia\Parsoid\Wt2Html\TokenHandlerPipeline; |
| 27 | |
| 28 | /** |
| 29 | * Generic attribute expansion handler. |
| 30 | */ |
| 31 | class AttributeExpander extends UniversalTokenHandler { |
| 32 | private const META_TYPE_MATCHER = '#(mw:(LanguageVariant|Transclusion|Param|Includes|Annotation/)(.*)$)#D'; |
| 33 | |
| 34 | /** |
| 35 | * Used for re-tokenizing attribute strings that need to be re-expanded |
| 36 | * @var PegTokenizer |
| 37 | */ |
| 38 | private $tokenizer; |
| 39 | |
| 40 | /** |
| 41 | * @param TokenHandlerPipeline $manager |
| 42 | * @param array $options |
| 43 | * - bool inTemplate Is this being invoked while processing a template? |
| 44 | * - bool expandTemplates Should we expand templates encountered here? |
| 45 | * - bool standalone Is this AttributeExpander used as part of a pipeline |
| 46 | * or is it being used standalone as an utility class? |
| 47 | */ |
| 48 | public function __construct( TokenHandlerPipeline $manager, array $options ) { |
| 49 | parent::__construct( $manager, $options ); |
| 50 | $this->tokenizer = new PegTokenizer( $manager->getEnv() ); |
| 51 | } |
| 52 | |
| 53 | private static function nlTkIndex( |
| 54 | bool $nlTkOkay, array $tokens, bool $atTopLevel |
| 55 | ): int { |
| 56 | // Moving this check here since it makes the |
| 57 | // callsite cleaner and simpler. |
| 58 | if ( $nlTkOkay ) { |
| 59 | return -1; |
| 60 | } |
| 61 | |
| 62 | // Check if we have a newline token in the attribute key/value token stream. |
| 63 | // However, newlines are acceptable inside a <*include*>..</*include*> directive |
| 64 | // since they are stripped out. |
| 65 | // |
| 66 | // let includeRE = !atTopLevel ? |
| 67 | // /(?:^|\s)mw:Includes\/NoInclude(\/.*)?(?:\s|$)/ : |
| 68 | // /(?:^|\s)mw:Includes\/(?:Only)?Include(?:Only)?(\/.*)?(?:\s|$)/; |
| 69 | // |
| 70 | // SSS FIXME: We cannot support this usage for <*include*> directives currently |
| 71 | // since they don't go through template encapsulation and don't have a data-mw |
| 72 | // format with "wt" and "transclusion" parts that we can use to just track bits |
| 73 | // of wikitext that don't have a DOM representation. |
| 74 | // |
| 75 | // So, for now, we just suppress all newlines contained within these directives. |
| 76 | $includeRE = '#(?:^|\s)mw:Includes/(?:No|Only)?Include(?:Only)?(/.*)?(?:\s|$)#D'; |
| 77 | $inInclude = false; |
| 78 | foreach ( $tokens as $i => $t ) { |
| 79 | if ( $t instanceof SelfclosingTagTk ) { |
| 80 | $type = $t->getAttributeV( 'typeof' ); |
| 81 | $typeMatch = []; |
| 82 | if ( $type && preg_match( $includeRE, $type, $typeMatch, PREG_UNMATCHED_AS_NULL ) ) { |
| 83 | $inInclude = !str_ends_with( $typeMatch[1] ?? '', '/End' ); |
| 84 | } |
| 85 | } elseif ( !$inInclude && $t instanceof NlTk ) { |
| 86 | // newline token outside <*include*> |
| 87 | return $i; |
| 88 | } |
| 89 | } |
| 90 | |
| 91 | return -1; |
| 92 | } |
| 93 | |
| 94 | /** |
| 95 | * @phpcs:ignore Generic.Files.LineLength.TooLong |
| 96 | * @return array{metaTokens: list{0?: SelfclosingTagTk}, preNLBuf: list<string|Token>, postNLBuf: list<string|Token>} |
| 97 | */ |
| 98 | private static function splitTokens( |
| 99 | Frame $frame, XMLTagTk $token, int $nlTkPos, array $tokens, bool $wrapTemplates |
| 100 | ): array { |
| 101 | $preNLBuf = []; |
| 102 | $postNLBuf = []; |
| 103 | $startMeta = null; |
| 104 | |
| 105 | // Split the token array around the first newline token. |
| 106 | $startMetaIndex = null; |
| 107 | foreach ( $tokens as $i => $t ) { |
| 108 | if ( $i === $nlTkPos ) { |
| 109 | // split here! |
| 110 | $postNLBuf = array_slice( $tokens, $i ); |
| 111 | break; |
| 112 | } else { |
| 113 | if ( $wrapTemplates && $t instanceof SelfclosingTagTk ) { |
| 114 | $type = $t->getAttributeV( 'typeof' ); |
| 115 | // We are interested in the last start meta tag. |
| 116 | // Everything before it is assumed to be closed. |
| 117 | $typeMatch = []; |
| 118 | if ( $type && |
| 119 | preg_match( self::META_TYPE_MATCHER, $type, $typeMatch ) && |
| 120 | !str_ends_with( $typeMatch[1], '/End' ) |
| 121 | ) { |
| 122 | $startMeta = $t; |
| 123 | $startMetaIndex = $i; |
| 124 | } |
| 125 | } |
| 126 | |
| 127 | // Use $i to make code robust if $tokens were not continugous |
| 128 | $preNLBuf[$i] = $t; |
| 129 | } |
| 130 | } |
| 131 | |
| 132 | // We split the token into pieces. |
| 133 | // Since we no longer know where this token now ends tsr-wise, |
| 134 | // set tsr->end to null |
| 135 | $token->dataParsoid->tsr->end = null; |
| 136 | $token->dataParsoid->getTemp()->attrSrc = ''; |
| 137 | |
| 138 | if ( $startMeta ) { |
| 139 | if ( count( $preNLBuf ) === 1 ) { |
| 140 | // Nothing to do since all real content (except the meta token) |
| 141 | // is after the newline. |
| 142 | return [ 'metaTokens' => [], 'preNLBuf' => [], 'postNLBuf' => $tokens ]; |
| 143 | } else { |
| 144 | // Clear $startMeta from $preNLBuf - setting to '' is sufficient. |
| 145 | $preNLBuf[$startMetaIndex] = ''; |
| 146 | |
| 147 | // Support template wrapping with the following steps: |
| 148 | // - Hoist the transclusion start-meta from the first line |
| 149 | // to before the token. |
| 150 | // - Update the start-meta tsr to that of the token. |
| 151 | // - Record the wikitext between the token and the transclusion |
| 152 | // as an unwrappedWT data-parsoid attribute of the start-meta. |
| 153 | $dp = $startMeta->dataParsoid; |
| 154 | $source = $dp->tsr->source ?? $frame->getSource(); |
| 155 | $dp->unwrappedWT = PHPUtils::safeSubstr( |
| 156 | $source->getSrcText(), $token->dataParsoid->tsr->start, |
| 157 | $dp->tsr->start - $token->dataParsoid->tsr->start ); |
| 158 | |
| 159 | // unwrappedWT will be added to the data-mw.parts array which makes |
| 160 | // this a multi-template-content-block. |
| 161 | // Record the first wikitext node of this block (required by html->wt serialization) |
| 162 | |
| 163 | // FIXME spec-compliant values would be upper-case, this is just a workaround |
| 164 | // for current PHP DOM implementation and could be removed in the future |
| 165 | $tokenName = mb_strtoupper( $token->getName() ); |
| 166 | |
| 167 | $dp->firstWikitextNode = isset( $token->dataParsoid->stx ) ? |
| 168 | $tokenName . '_' . $token->dataParsoid->stx : $tokenName; |
| 169 | |
| 170 | // Update tsr->start only. Unless the end-meta token is moved as well, |
| 171 | // updating tsr->end can introduce bugs in cases like: |
| 172 | // |
| 173 | // {| |
| 174 | // |{{singlechart|Australia|93|artist=Madonna|album=Girls Gone Wild}}|x |
| 175 | // |} |
| 176 | // |
| 177 | // which can then cause dirty diffs (the "|" before the x gets dropped). |
| 178 | $dp->tsr->start = $token->dataParsoid->tsr->start; |
| 179 | $metaTokens = [ $startMeta ]; |
| 180 | |
| 181 | return [ 'metaTokens' => $metaTokens, 'preNLBuf' => $preNLBuf, 'postNLBuf' => $postNLBuf ]; |
| 182 | } |
| 183 | } else { |
| 184 | return [ 'metaTokens' => [], 'preNLBuf' => $tokens, 'postNLBuf' => [] ]; |
| 185 | } |
| 186 | } |
| 187 | |
| 188 | /** |
| 189 | * This helper method strips all meta tags introduced by |
| 190 | * transclusions, etc. and returns the content. |
| 191 | * |
| 192 | * @param Env $env |
| 193 | * @param array $tokens |
| 194 | * @param bool $wrapTemplates |
| 195 | * |
| 196 | * @return array{hasGeneratedContent: bool, annotationType: list<string>, value: list<Token|string>} |
| 197 | */ |
| 198 | private static function stripMetaTags( |
| 199 | Env $env, array $tokens, bool $wrapTemplates, bool $inTableCellContext = false |
| 200 | ): array { |
| 201 | $buf = []; |
| 202 | $hasGeneratedContent = false; |
| 203 | $cellAttrTerminatorSeen = false; |
| 204 | $annotationType = []; |
| 205 | |
| 206 | foreach ( $tokens as $t ) { |
| 207 | if ( $t instanceof TagTk || $t instanceof SelfclosingTagTk ) { |
| 208 | // Take advantage of this iteration of `tokens` to seek out |
| 209 | // document fragments. They're an indication that an attribute |
| 210 | // value wasn't present as literal text in the input and the |
| 211 | // token should be annotated with "mw:ExpandedAttrs". |
| 212 | if ( TokenUtils::hasDOMFragmentType( $t ) ) { |
| 213 | $hasGeneratedContent = true; |
| 214 | } |
| 215 | |
| 216 | if ( $inTableCellContext ) { |
| 217 | // Images, links, etc. terminate table cell attribute processing. |
| 218 | // If so, record info so TableFixups can cleanly convert this attribute |
| 219 | // back to content and handle any additional fixups resulting from that. |
| 220 | $rel = $t->getAttributeV( 'rel' ) ?? ''; |
| 221 | $typeof = $t->getAttributeV( 'typeof' ) ?? ''; |
| 222 | if ( preg_match( '#\bmw:File($|/)\b#', $typeof ) || |
| 223 | preg_match( WTUtils::WIKILINK_SYNTAX_CONSTRUCTS_REGEXP, $rel ) |
| 224 | ) { |
| 225 | $cellAttrTerminatorSeen = true; |
| 226 | } |
| 227 | } |
| 228 | |
| 229 | if ( $wrapTemplates ) { |
| 230 | // Strip all meta tags. |
| 231 | $type = $t->getAttributeV( 'typeof' ); |
| 232 | $typeMatch = []; |
| 233 | if ( $type && preg_match( self::META_TYPE_MATCHER, $type, $typeMatch ) ) { |
| 234 | if ( !str_ends_with( $typeMatch[1], '/End' ) ) { |
| 235 | $hasGeneratedContent = true; |
| 236 | } |
| 237 | $groups = []; |
| 238 | if ( preg_match( WTUtils::ANNOTATION_META_TYPE_REGEXP, $type, $groups ) ) { |
| 239 | $annotationType[] = $groups[1]; |
| 240 | } |
| 241 | } else { |
| 242 | $buf[] = $t; |
| 243 | continue; |
| 244 | } |
| 245 | } |
| 246 | |
| 247 | if ( $t->getName() !== 'meta' ) { |
| 248 | // Dont strip token if it is not a meta-tag |
| 249 | $buf[] = $t; |
| 250 | } |
| 251 | } else { |
| 252 | $buf[] = $t; |
| 253 | } |
| 254 | } |
| 255 | |
| 256 | return [ |
| 257 | 'hasGeneratedContent' => $hasGeneratedContent, |
| 258 | 'cellAttrTerminatorSeen' => $cellAttrTerminatorSeen, |
| 259 | 'annotationType' => $annotationType, |
| 260 | 'value' => $buf |
| 261 | ]; |
| 262 | } |
| 263 | |
| 264 | /** |
| 265 | * This receives tokens from the PegTokenizer whose input itself |
| 266 | * is preprocessed wikitext. So, templates and include directives |
| 267 | * have all been processed and comments have been dropped. |
| 268 | * Anything left behind (directives, template tokens, template args) |
| 269 | * should just be converted to a plain string. |
| 270 | * |
| 271 | * @param mixed $a |
| 272 | */ |
| 273 | private static function tplToksToString( $a ): string { |
| 274 | if ( !is_array( $a ) ) { |
| 275 | return $a; |
| 276 | } |
| 277 | $buf = ""; |
| 278 | foreach ( $a as $t ) { |
| 279 | if ( is_string( $t ) ) { |
| 280 | $buf .= $t; |
| 281 | } else { |
| 282 | $buf .= $t->dataParsoid->src ?? ''; /* drop it if dp->src is missing */ |
| 283 | } |
| 284 | } |
| 285 | return $buf; |
| 286 | } |
| 287 | |
| 288 | /** |
| 289 | * Callback for attribute expansion in AttributeTransformManager |
| 290 | * @param XMLTagTk $token |
| 291 | * @param KV[] $expandedAttrs |
| 292 | * @return array<string|Token> |
| 293 | */ |
| 294 | private function buildExpandedAttrs( XMLTagTk $token, array $expandedAttrs ): array { |
| 295 | // If we're not in a template, we'll be doing template wrapping in dom |
| 296 | // post-processing (same conditional there), so take care of meta markers |
| 297 | // found while processing tokens. |
| 298 | $wrapTemplates = !$this->options['inTemplate']; |
| 299 | $env = $this->manager->getEnv(); |
| 300 | $metaTokens = []; |
| 301 | $postNLToks = []; |
| 302 | $tmpDataMW = null; |
| 303 | $oldAttrs = $token->attribs; |
| 304 | $tokenName = $token->getName(); |
| 305 | // Build newAttrs lazily (on-demand) to avoid creating |
| 306 | // objects in the common case where nothing of significance |
| 307 | // happens in this code. |
| 308 | $newAttrs = null; |
| 309 | // FIXME: td/th/caption need different handling. |
| 310 | // For now, we are limiting this to table & tr tags because |
| 311 | // the code below uses 'table_attributes' to reparse the string |
| 312 | // which is only valid for table & tr tokens. The fix for td/th/caption |
| 313 | // may be as simple as using the 'row_syntax_table_args' rule. To be |
| 314 | // investigated and fixed. |
| 315 | $nlTkOkay = TokenUtils::isHTMLTag( $token ) || |
| 316 | ( $tokenName !== 'table' && $tokenName !== 'tr' ); |
| 317 | $annotationTypes = []; |
| 318 | |
| 319 | // Identify attributes that were generated in full or in part using templates |
| 320 | foreach ( $oldAttrs as $i => $oldA ) { |
| 321 | $expandedA = $expandedAttrs[$i]; |
| 322 | |
| 323 | // Preserve the key and value source, if available. |
| 324 | // But, if 'oldA' wasn't cloned, expandedA will be the same as 'oldA'. |
| 325 | if ( $oldA !== $expandedA ) { |
| 326 | $expandedA->ksrc = $oldA->ksrc; |
| 327 | $expandedA->vsrc = $oldA->vsrc; |
| 328 | $expandedA->srcOffsets = $oldA->srcOffsets; |
| 329 | } |
| 330 | |
| 331 | // Deal with two template-expansion scenarios for the attribute key (not value) |
| 332 | // |
| 333 | // 1. We have a template that generates multiple attributes of this token |
| 334 | // as well as content after the token. |
| 335 | // Ex: infobox templates from aircraft, ship, and other pages |
| 336 | // See enwiki:Boeing_757 |
| 337 | // |
| 338 | // - Split the expanded tokens into multiple lines. |
| 339 | // - Expanded attributes associated with the token are retained in the |
| 340 | // first line before a NlTk. |
| 341 | // - Content tokens after the NlTk are moved to subsequent lines. |
| 342 | // - The meta tags are hoisted before the original token to make sure |
| 343 | // that the entire token and following content is encapsulated as a unit. |
| 344 | // |
| 345 | // 2. We have a template that only generates multiple attributes of this |
| 346 | // token. In that case, we strip all template meta tags from the expanded |
| 347 | // tokens and assign it a mw:ExpandedAttrs type with orig/expanded |
| 348 | // values in data-mw. |
| 349 | // |
| 350 | // Reparse-KV-string scenario with templated attributes: |
| 351 | // ----------------------------------------------------- |
| 352 | // In either scenario above, we need additional special handling if the |
| 353 | // template generates one or more k=v style strings: |
| 354 | // <div {{1x|1=style='color:red''}}></div> |
| 355 | // <div {{1x|1=style='color:red' title='boo'}}></div> |
| 356 | // |
| 357 | // Real use case: Template {{ligne grise}} on frwp. |
| 358 | // |
| 359 | // To support this, we utilize the following hack. If we got a string of the |
| 360 | // form "k=v" and our orig-v was "", we convert the token array to a string |
| 361 | // and retokenize it to extract one or more attributes. |
| 362 | // |
| 363 | // But, we won't support scenarios like this: |
| 364 | // {| title={{1x|1='name' style='color:red;'\n|-\n|foo}}\n|} |
| 365 | // Here, part of one attribute and additional complete attribute strings |
| 366 | // need reparsing, and that isn't a use case that is worth more complexity here. |
| 367 | // |
| 368 | // FIXME: |
| 369 | // ------ |
| 370 | // 1. It is not possible for multiple instances of scenario 1 to be triggered |
| 371 | // for the same token. So, I am not bothering trying to test and deal with it. |
| 372 | // |
| 373 | // 2. We trigger the Reparse-KV-string scenario only for attribute keys, |
| 374 | // since it isn't possible for attribute values to require this reparsing. |
| 375 | // However, it is possible to come up with scenarios where a template |
| 376 | // returns the value for one attribute and additional k=v strings for newer |
| 377 | // attributes. We don't support that scenario, but don't even test for it. |
| 378 | // |
| 379 | // Reparse-KV-string scenario with non-string attributes: |
| 380 | // ------------------------------------------------------ |
| 381 | // This is only going to be the case with table wikitext that has special syntax |
| 382 | // for attribute strings. |
| 383 | // |
| 384 | // {| <div>a</div> style='border:1px solid black;' |
| 385 | // |- <div>b</div> style='border:1px dotted blue;' |
| 386 | // | <div>c</div> style='color:red;' |
| 387 | // |} |
| 388 | // |
| 389 | // In wikitext like the above, the PEG tokenizer doesn't recognize these as |
| 390 | // valid attributes (the templated attribute scenario is a special case) and |
| 391 | // orig-v will be "". So, the same strategy as above is applied here as well. |
| 392 | |
| 393 | $expandedK = $origK = $expandedA->k; |
| 394 | $expandedV = $origV = $expandedA->v; |
| 395 | $updatedK = null; |
| 396 | $updatedV = null; |
| 397 | $reparsedKV = false; |
| 398 | $keyUsesMixedAttrContentTpl = false; |
| 399 | $valUsesMixedAttrContentTpl = false; |
| 400 | |
| 401 | if ( $expandedK ) { |
| 402 | // FIXME: We should get rid of these array/string/non-string checks |
| 403 | // and probably use appropriately-named flags to convey type information. |
| 404 | if ( is_array( $oldA->k ) ) { |
| 405 | if ( !is_array( $expandedK ) ) { |
| 406 | throw new UnreachableException( "expandedK: expected array. Found: " . |
| 407 | PHPUtils::jsonEncode( $expandedK ) ); |
| 408 | } |
| 409 | |
| 410 | $nlTkPos = self::nlTkIndex( $nlTkOkay, $expandedK, $wrapTemplates ); |
| 411 | if ( $nlTkPos !== -1 ) { |
| 412 | // Scenario 1 from the documentation comment above. |
| 413 | $keyUsesMixedAttrContentTpl = true; |
| 414 | $updatedK = self::splitTokens( |
| 415 | $this->manager->getFrame(), $token, $nlTkPos, $expandedK, $wrapTemplates |
| 416 | ); |
| 417 | $expandedK = $updatedK['preNLBuf']; |
| 418 | $postNLToks = $updatedK['postNLBuf']; |
| 419 | $metaTokens = $updatedK['metaTokens']; |
| 420 | // We split up this attribute's key into pieces. |
| 421 | if ( $expandedA->srcOffsets->key ) { |
| 422 | $expandedA->srcOffsets->key->end = null; |
| 423 | $expandedA->ksrc = null; |
| 424 | } |
| 425 | } else { |
| 426 | // Maybe scenario 2 from the documentation comment above. |
| 427 | $updatedK = self::stripMetaTags( |
| 428 | $env, $expandedK, $wrapTemplates, |
| 429 | ( $tokenName === 'td' || $tokenName === 'th' ) |
| 430 | ); |
| 431 | PHPUtils::pushArray( $annotationTypes, $updatedK['annotationType'] ); |
| 432 | $expandedK = $updatedK['value']; |
| 433 | // @phan-suppress-next-line PhanTypeInvalidDimOffset |
| 434 | if ( $updatedK['cellAttrTerminatorSeen'] ) { |
| 435 | $token->dataParsoid->getTemp()->cellAttrTerminatorSeen = true; |
| 436 | } |
| 437 | } |
| 438 | |
| 439 | $expandedA->k = $expandedK; |
| 440 | |
| 441 | // Check if we need to deal with the Reparse-KV-string scenario. |
| 442 | // (See documentation comment above.) |
| 443 | // |
| 444 | // Don't incorrectly reparse the kv string for parser functions. |
| 445 | // Ex: "#ifexpr" parser function expects the "=" equality operator. |
| 446 | // We encounter those in "standalone" mode (used to expand |
| 447 | // templated template targets). |
| 448 | if ( $expandedA->v === '' && empty( $this->options['standalone'] ) ) { |
| 449 | // Extract a parsable string from the token array. |
| 450 | // Trim whitespace to ensure tokenizer isn't tripped up |
| 451 | // by the presence of unnecessary whitespace. |
| 452 | $kStr = trim( TokenUtils::tokensToString( $expandedK, false, [ |
| 453 | // These tokens haven't been expanded to DOM yet |
| 454 | // so unpacking them here is justifiable |
| 455 | 'unpackDOMFragments' => true, |
| 456 | ] ) ); |
| 457 | $rule = $nlTkOkay ? 'generic_newline_attributes' : 'table_attributes'; |
| 458 | $kvs = str_contains( $kStr, '=' ) ? |
| 459 | $this->tokenizer->tokenizeAs( $kStr, $rule, /* sol */true ) : null; |
| 460 | if ( $kvs ) { |
| 461 | // At this point, templates should have been expanded. |
| 462 | // Returning a template token here probably means that |
| 463 | // when we just converted to string and reparsed, we failed |
| 464 | // to expand the template. This can be particularly bad |
| 465 | // when we make iterative calls to expand template names. |
| 466 | // So, give up template expansion and convert them to strings. |
| 467 | foreach ( $kvs as $kv ) { |
| 468 | $kv->k = self::tplToksToString( $kv->k ); |
| 469 | $kv->v = self::tplToksToString( $kv->v ); |
| 470 | |
| 471 | // $kStr is based on running tokensToString on $expandedK. |
| 472 | // So, $kStr might have dropped HTML tags, etc. Given that, |
| 473 | // we can no longer reliably compute offsets for these |
| 474 | // new key/value pairs. We could try to be more smart here, |
| 475 | // but it is not worth the complexity. |
| 476 | $kv->srcOffsets = null; |
| 477 | } |
| 478 | // SSS FIXME: Collect all keys here, not just the first key |
| 479 | // i.e. in a string like {{1x|1=id='v1' title='foo' style='..'}} |
| 480 | // that string is setting attributes for [id, title, style], not just id. |
| 481 | // |
| 482 | // That requires the ability for the data-mw.attribs[i].txt to be an array. |
| 483 | // However, the spec at [[mw:Specs/HTML#Generated_attributes_of_HTML_tags]] |
| 484 | // says: |
| 485 | // |
| 486 | // "This spec also assumes that a template can only |
| 487 | // generate one attribute rather than multiple attributes." |
| 488 | // |
| 489 | // So, revision of the spec is another FIXME at which point this code can |
| 490 | // be updated to reflect the revised spec. |
| 491 | $expandedK = $kvs[0]->k; |
| 492 | $reparsedKV = true; |
| 493 | if ( !$newAttrs ) { |
| 494 | $newAttrs = $i === 0 ? [] : array_slice( $expandedAttrs, 0, $i ); |
| 495 | } |
| 496 | PHPUtils::pushArray( $newAttrs, $kvs ); |
| 497 | } |
| 498 | } |
| 499 | } |
| 500 | |
| 501 | // We have a potentially expanded value. |
| 502 | // Check if the value came from a template/extension expansion. |
| 503 | if ( is_string( $expandedK ) && !str_starts_with( $expandedK, 'mw:' ) |
| 504 | && is_array( $oldA->v ) |
| 505 | ) { |
| 506 | $nlTkPos = self::nlTkIndex( $nlTkOkay, $expandedV, $wrapTemplates ); |
| 507 | if ( $nlTkPos !== -1 ) { |
| 508 | // Scenario 1 from the documentation comment above. |
| 509 | $valUsesMixedAttrContentTpl = true; |
| 510 | $updatedV = self::splitTokens( |
| 511 | $this->manager->getFrame(), $token, $nlTkPos, |
| 512 | $expandedV, $wrapTemplates |
| 513 | ); |
| 514 | $expandedV = $updatedV['preNLBuf']; |
| 515 | $postNLToks = $updatedV['postNLBuf']; |
| 516 | $metaTokens = $updatedV['metaTokens']; |
| 517 | // We split up this attribute's value into pieces. |
| 518 | if ( $expandedA->srcOffsets->value ) { |
| 519 | $expandedA->srcOffsets->value->end = null; |
| 520 | $expandedA->vsrc = null; |
| 521 | } |
| 522 | } else { |
| 523 | // Maybe scenario 2 from the documentation comment above. |
| 524 | $updatedV = self::stripMetaTags( |
| 525 | $env, $expandedV, $wrapTemplates, |
| 526 | ( $tokenName === 'td' || $tokenName === 'th' ) |
| 527 | ); |
| 528 | PHPUtils::pushArray( $annotationTypes, $updatedV['annotationType'] ); |
| 529 | $expandedV = $updatedV['value']; |
| 530 | // @phan-suppress-next-line PhanTypeInvalidDimOffset |
| 531 | if ( $updatedV['cellAttrTerminatorSeen'] ) { |
| 532 | $token->dataParsoid->getTemp()->cellAttrTerminatorSeen = true; |
| 533 | } |
| 534 | } |
| 535 | $expandedA->v = $expandedV; |
| 536 | } |
| 537 | |
| 538 | // Update data-mw to account for templated attributes. |
| 539 | // For editability, set HTML property. |
| 540 | if ( !empty( $updatedK['hasGeneratedContent'] ) || |
| 541 | !empty( $updatedV['hasGeneratedContent'] ) || |
| 542 | ( $reparsedKV && count( $metaTokens ) > 0 ) |
| 543 | ) { |
| 544 | $key = TokenUtils::tokensToString( $expandedK ); |
| 545 | if ( !$tmpDataMW ) { |
| 546 | $tmpDataMW = []; |
| 547 | } |
| 548 | |
| 549 | // For the $(key|val)UsesMixedAttrContentTpl checks below, |
| 550 | // it is incorrect to assign the HTML for the original wikitext |
| 551 | // string since the content part will get duplicated in both |
| 552 | // this data-mw and in the actual body of the table (for example) |
| 553 | // and cause bugs like T249740. |
| 554 | // |
| 555 | // So, in this case, we assign just the key/value part of the HTML |
| 556 | // ($expandedA->k or $expandedA->v), but we mark it uneditable |
| 557 | // because we cannot really edit just the key/value of the attribute |
| 558 | // on its own because it is only a part of the template's output. |
| 559 | if ( $reparsedKV ) { |
| 560 | // If we encountered a reparse-KV-string scenario, |
| 561 | // we set the value's HTML to [] since we can edit |
| 562 | // the transclusion either via the key's HTML or the |
| 563 | // value's HTML, but not both. |
| 564 | $keyHTML = $keyUsesMixedAttrContentTpl ? $expandedA->k : $origK; |
| 565 | $valHTML = []; |
| 566 | } else { |
| 567 | Assert::invariant( !$keyUsesMixedAttrContentTpl, |
| 568 | "If reparseKV was false, and we had a mixed attr-content template, " . |
| 569 | "we should have landed in the valUsesMixedAttrContentTpl codepath." ); |
| 570 | $keyHTML = empty( $updatedK['hasGeneratedContent'] ) ? null : $origK; |
| 571 | $valHTML = $valUsesMixedAttrContentTpl ? $expandedA->v : $origV; |
| 572 | } |
| 573 | |
| 574 | // FIXME: Ideally we would have called them ktext, khtml, vhtml |
| 575 | // since in the serialized data-mw, the "k" and "v" key strings are dropped. |
| 576 | // [{ "ktxt":..., "khtml":... }, { "vhtml":... }] |
| 577 | // is clearer and less confusing than |
| 578 | // [{ "txt":..., "html":... }, { "html":... }] |
| 579 | $tmpDataMW[$key] = [ |
| 580 | // @phan-suppress-next-line PhanCoalescingNeverNullInLoop $expandedA is nullable |
| 581 | 'k' => [ 'txt' => $key, 'srcOffsets' => $expandedA->srcOffsets->key ?? null ], |
| 582 | // FIXME: Why is 'txt' missing? Why are we not checking for [] ? |
| 583 | // @phan-suppress-next-line PhanCoalescingNeverNullInLoop $expandedA is nullable |
| 584 | 'v' => [ 'html' => $valHTML, 'srcOffsets' => $expandedA->srcOffsets->value ?? null ] |
| 585 | ]; |
| 586 | |
| 587 | if ( $keyHTML !== null ) { |
| 588 | $tmpDataMW[$key]['k']['html'] = $keyHTML; |
| 589 | } |
| 590 | if ( $keyUsesMixedAttrContentTpl ) { |
| 591 | $tmpDataMW[$key]['k']['uneditable'] = true; |
| 592 | } |
| 593 | if ( $valUsesMixedAttrContentTpl ) { |
| 594 | $tmpDataMW[$key]['v']['uneditable'] = true; |
| 595 | } |
| 596 | } |
| 597 | } |
| 598 | |
| 599 | // Update newAttrs |
| 600 | if ( $newAttrs && !$reparsedKV ) { |
| 601 | $newAttrs[] = $expandedA; |
| 602 | } |
| 603 | } |
| 604 | |
| 605 | $token->attribs = $newAttrs ?? $expandedAttrs; |
| 606 | |
| 607 | // If the token already has an about, it already has transclusion/extension |
| 608 | // wrapping. No need to record information about templated attributes in addition. |
| 609 | // |
| 610 | // FIXME: If there is a real use case for extension attributes getting templated, |
| 611 | // this check can be relaxed to allow that. |
| 612 | // https://gerrit.wikimedia.org/r/#/c/65575 has some reference code that can be used then. |
| 613 | |
| 614 | if ( !$token->getAttributeV( 'about' ) && $tmpDataMW && count( $tmpDataMW ) > 0 ) { |
| 615 | // Flatten k-v pairs. |
| 616 | $vals = []; |
| 617 | foreach ( $tmpDataMW as $obj ) { |
| 618 | $vals[] = $obj['k']; |
| 619 | $vals[] = $obj['v']; |
| 620 | } |
| 621 | |
| 622 | // Clone the vals since they'll be passed to another pipeline |
| 623 | // for expanding, which may destructively mutate them in the process. |
| 624 | // |
| 625 | // This is a problem since subsequent handlers to the |
| 626 | // AttributeExpander may interact with the original tokens still |
| 627 | // present as attributes of `token`. |
| 628 | // |
| 629 | // For example, while treebuilding, the object holding dataParsoid |
| 630 | // of a token is reused as the data-parsoid attribute of the |
| 631 | // corresonding node. Thus, when we get to the DOM cleanup pass, |
| 632 | // unsetting properties changes the token as well. This was |
| 633 | // the issue when an "href" was expanded and then the |
| 634 | // ExternalLinkHandler tried to call tokensToString on it, |
| 635 | // resulting in a transcluded entity missing its src (which, by the way, |
| 636 | // had already been clobered by WrapTemplates, similar to T214241). |
| 637 | // |
| 638 | // The general principle here being, don't share tokens between |
| 639 | // pipelines. |
| 640 | $vals = Utils::cloneArray( $vals ); |
| 641 | |
| 642 | // Expand all token arrays to DOM. |
| 643 | $eVals = PipelineUtils::expandAttrValuesToDOM( |
| 644 | $this->env, $this->manager->getFrame(), $vals, |
| 645 | $this->options['expandTemplates'], |
| 646 | $this->options['inTemplate'] |
| 647 | ); |
| 648 | |
| 649 | // Rebuild flattened k-v pairs. |
| 650 | $expAttrs = []; |
| 651 | for ( $j = 0; $j < count( $eVals ); $j += 2 ) { |
| 652 | $expAttrs[] = new DataMwAttrib( $eVals[$j], $eVals[$j + 1] ); |
| 653 | } |
| 654 | |
| 655 | // Mark token as having expanded attrs. |
| 656 | // |
| 657 | // Template tokens are omitted because the attribute expander is |
| 658 | // just being used to resolve the template target. |
| 659 | if ( $tokenName !== 'template' ) { |
| 660 | $token->addAttribute( 'about', $this->env->newAboutId() ); |
| 661 | $token->addSpaceSeparatedAttribute( 'typeof', 'mw:ExpandedAttrs' ); |
| 662 | foreach ( $annotationTypes as $annotationType ) { |
| 663 | $token->addSpaceSeparatedAttribute( 'typeof', 'mw:Annotation/' . $annotationType ); |
| 664 | } |
| 665 | $token->dataMw = new DataMw( [ 'attribs' => $expAttrs ] ); |
| 666 | } |
| 667 | } |
| 668 | |
| 669 | return array_merge( $metaTokens, [ $token ], $postNLToks ); |
| 670 | } |
| 671 | |
| 672 | /** |
| 673 | * Processes any attribute keys and values that are not simple strings. |
| 674 | * (Ex: Templated styles) |
| 675 | * |
| 676 | * @param XMLTagTk $token Token whose attrs being expanded. |
| 677 | * @return ?array<string|Token> |
| 678 | */ |
| 679 | private function processComplexAttributes( XMLTagTk $token ): ?array { |
| 680 | $expandedAttrs = AttributeTransformManager::process( |
| 681 | $this->manager->getFrame(), |
| 682 | [ |
| 683 | 'expandTemplates' => $this->options['expandTemplates'], |
| 684 | 'inTemplate' => $this->options['inTemplate'] |
| 685 | ], |
| 686 | $token->attribs |
| 687 | ); |
| 688 | |
| 689 | // detection for linting ( |
| 690 | if ( $token->getName() === 'urllink' || $token->getName() === 'extlink' ) { |
| 691 | /** @var null|array<Token> $hrefAttr */ |
| 692 | $hrefAttr = $token->getAttributeV( "href" ); |
| 693 | if ( TokenUtils::hasTemplateToken( $hrefAttr ) ) { |
| 694 | $token->dataParsoid->getTemp()->linkContainsTemplate = true; |
| 695 | } |
| 696 | } |
| 697 | |
| 698 | // null signifies unmodified token |
| 699 | return $expandedAttrs ? $this->buildExpandedAttrs( $token, $expandedAttrs ) : null; |
| 700 | } |
| 701 | |
| 702 | /** |
| 703 | * Expand the first attribute of the token -- usually needed to support |
| 704 | * tempate tokens where the template target itself is a complex attribute. |
| 705 | * |
| 706 | * @param XMLTagTK $token Token whose first attribute is being expanded. |
| 707 | * @return ?array<string|Token> |
| 708 | */ |
| 709 | public function expandFirstAttribute( XMLTagTk $token ): ?array { |
| 710 | $expandedAttrs = AttributeTransformManager::process( |
| 711 | $this->manager->getFrame(), |
| 712 | [ |
| 713 | 'expandTemplates' => $this->options['expandTemplates'], |
| 714 | 'inTemplate' => $this->options['inTemplate'] |
| 715 | ], |
| 716 | [ $token->attribs[0] ] |
| 717 | ); |
| 718 | if ( $expandedAttrs ) { |
| 719 | return $this->buildExpandedAttrs( |
| 720 | $token, |
| 721 | array_replace( $token->attribs, [ 0 => $expandedAttrs[0] ] ) |
| 722 | ); |
| 723 | } else { |
| 724 | // null signifies unmodified token |
| 725 | return null; |
| 726 | } |
| 727 | } |
| 728 | |
| 729 | /** @inheritDoc */ |
| 730 | public function onCompoundTk( CompoundTk $ctk, TokenHandler $tokensHandler ): ?array { |
| 731 | if ( $ctk instanceof EmptyLineTk ) { |
| 732 | return null; |
| 733 | } else { |
| 734 | // default handling: in this case throws exception! |
| 735 | return parent::onCompoundTk( $ctk, $tokensHandler ); |
| 736 | } |
| 737 | } |
| 738 | |
| 739 | /** |
| 740 | * For tokens that might have complex attributes, this handler processes / expands them. |
| 741 | * (Ex: Templated styles) |
| 742 | * |
| 743 | * @inheritDoc |
| 744 | */ |
| 745 | public function onAny( $token ): ?array { |
| 746 | if ( |
| 747 | !( $token instanceof TagTk || $token instanceof SelfclosingTagTk ) || |
| 748 | !count( $token->attribs ) |
| 749 | ) { |
| 750 | return null; |
| 751 | } |
| 752 | |
| 753 | $name = $token->getName(); |
| 754 | $typeOf = $token->getAttributeV( 'typeof' ) ?? ''; |
| 755 | |
| 756 | if ( |
| 757 | // Do not process dom-fragment tokens: a separate handler deals with them. |
| 758 | $name === 'mw:dom-fragment-token' || |
| 759 | // Parsoid generated metas don't need expansion |
| 760 | ( $name === 'meta' && |
| 761 | preg_match( '/mw:(Placeholder|Transclusion|Param|Includes)/', $typeOf ) ) |
| 762 | ) { |
| 763 | return null; |
| 764 | } |
| 765 | |
| 766 | Assert::invariant( |
| 767 | !str_contains( $typeOf, 'mw:ExpandedAttrs' ), |
| 768 | "Expanding an already expanded token, that's a no-no." |
| 769 | ); |
| 770 | |
| 771 | return $this->processComplexAttributes( $token ); |
| 772 | } |
| 773 | } |