Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
0.00% |
0 / 191 |
|
0.00% |
0 / 11 |
CRAP | |
0.00% |
0 / 1 |
| TokenStreamPatcher | |
0.00% |
0 / 191 |
|
0.00% |
0 / 11 |
7482 | |
0.00% |
0 / 1 |
| __construct | |
0.00% |
0 / 4 |
|
0.00% |
0 / 1 |
2 | |||
| resetState | |
0.00% |
0 / 2 |
|
0.00% |
0 / 1 |
6 | |||
| reset | |
0.00% |
0 / 5 |
|
0.00% |
0 / 1 |
2 | |||
| onNewline | |
0.00% |
0 / 14 |
|
0.00% |
0 / 1 |
30 | |||
| onEnd | |
0.00% |
0 / 3 |
|
0.00% |
0 / 1 |
2 | |||
| clearSOL | |
0.00% |
0 / 2 |
|
0.00% |
0 / 1 |
2 | |||
| reprocessTokens | |
0.00% |
0 / 11 |
|
0.00% |
0 / 1 |
6 | |||
| convertTokenToString | |
0.00% |
0 / 22 |
|
0.00% |
0 / 1 |
182 | |||
| onAny | |
0.00% |
0 / 4 |
|
0.00% |
0 / 1 |
6 | |||
| handleT2529Hack | |
0.00% |
0 / 8 |
|
0.00% |
0 / 1 |
56 | |||
| onAnyInternal | |
0.00% |
0 / 116 |
|
0.00% |
0 / 1 |
2652 | |||
| 1 | <?php |
| 2 | declare( strict_types = 1 ); |
| 3 | |
| 4 | namespace Wikimedia\Parsoid\Wt2Html\TT; |
| 5 | |
| 6 | use Wikimedia\Parsoid\NodeData\DataParsoid; |
| 7 | use Wikimedia\Parsoid\Tokens\EndTagTk; |
| 8 | use Wikimedia\Parsoid\Tokens\EOFTk; |
| 9 | use Wikimedia\Parsoid\Tokens\KV; |
| 10 | use Wikimedia\Parsoid\Tokens\NlTk; |
| 11 | use Wikimedia\Parsoid\Tokens\SelfclosingTagTk; |
| 12 | use Wikimedia\Parsoid\Tokens\TagTk; |
| 13 | use Wikimedia\Parsoid\Tokens\Token; |
| 14 | use Wikimedia\Parsoid\Utils\PHPUtils; |
| 15 | use Wikimedia\Parsoid\Utils\TokenUtils; |
| 16 | use Wikimedia\Parsoid\Wt2Html\PegTokenizer; |
| 17 | use Wikimedia\Parsoid\Wt2Html\TokenTransformManager; |
| 18 | |
| 19 | /** |
| 20 | * This class is an attempt to fixup the token stream to reparse strings |
| 21 | * as tokens that failed to parse in the tokenizer because of SOL or |
| 22 | * other constraints OR because tags were being constructed in pieces |
| 23 | * or whatever. |
| 24 | * |
| 25 | * This is a pure hack to improve compatibility with the core parser |
| 26 | * given that we dont have a preprocessor. This will be a grab-bag of |
| 27 | * heuristics and tricks to handle different scenarios. |
| 28 | */ |
| 29 | class TokenStreamPatcher extends TokenHandler { |
| 30 | private PegTokenizer $tokenizer; |
| 31 | |
| 32 | /** @var int|null */ |
| 33 | private $srcOffset; |
| 34 | |
| 35 | private bool $sol; |
| 36 | |
| 37 | private array $tokenBuf; |
| 38 | private int $wikiTableNesting; |
| 39 | /** True only for top-level & attribute value pipelines */ |
| 40 | private bool $inIndependentParse; |
| 41 | |
| 42 | /** @var Token|null */ |
| 43 | private $lastConvertedTableCellToken; |
| 44 | |
| 45 | /** @var SelfclosingTagTk|null */ |
| 46 | private $tplStartToken = null; |
| 47 | |
| 48 | /** @var NlTk|null */ |
| 49 | private $discardableNlTk = null; |
| 50 | |
| 51 | public function __construct( TokenTransformManager $manager, array $options ) { |
| 52 | $newOptions = [ 'tsp' => true ] + $options; |
| 53 | parent::__construct( $manager, $newOptions ); |
| 54 | $this->tokenizer = new PegTokenizer( $this->env ); |
| 55 | $this->reset(); |
| 56 | } |
| 57 | |
| 58 | /** |
| 59 | * Resets any internal state for this token handler. |
| 60 | * |
| 61 | * @param array $parseOpts |
| 62 | */ |
| 63 | public function resetState( array $parseOpts ): void { |
| 64 | parent::resetState( $parseOpts ); |
| 65 | $this->inIndependentParse = $this->atTopLevel || isset( $this->options['attrExpansion'] ); |
| 66 | } |
| 67 | |
| 68 | private function reset() { |
| 69 | $this->srcOffset = 0; |
| 70 | $this->sol = true; |
| 71 | $this->tokenBuf = []; |
| 72 | $this->wikiTableNesting = 0; |
| 73 | // This marker tries to track the most recent table-cell token (td/th) |
| 74 | // that was converted to string. For those, we want to get rid |
| 75 | // of their corresponding mw:TSRMarker meta tag. |
| 76 | // |
| 77 | // This marker is set when we convert a td/th token to string |
| 78 | // |
| 79 | // This marker is cleared in one of the following scenarios: |
| 80 | // 1. When we clear a mw:TSRMarker corresponding to the token set earlier |
| 81 | // 2. When we change table nesting |
| 82 | // 3. When we hit a tr/td/th/caption token that wasn't converted to string |
| 83 | $this->lastConvertedTableCellToken = null; |
| 84 | } |
| 85 | |
| 86 | /** |
| 87 | * @inheritDoc |
| 88 | */ |
| 89 | public function onNewline( NlTk $token ): ?TokenHandlerResult { |
| 90 | $self = $this; |
| 91 | $this->env->log( 'trace/tsp', $this->pipelineId, |
| 92 | static function () use ( $self, $token ) { |
| 93 | return "(indep=" . ( $self->inIndependentParse ? "yes" : "no " ) . |
| 94 | ";sol=" . ( $self->sol ? "yes" : "no " ) . |
| 95 | PHPUtils::jsonEncode( $token ); |
| 96 | } |
| 97 | ); |
| 98 | $this->srcOffset = $token->dataParsoid->tsr->end ?? null; |
| 99 | if ( $this->sol && $this->tplStartToken ) { |
| 100 | // When using core preprocessor, start-of-line start is forced by |
| 101 | // inserting a newline in certain cases (the "T2529 hack"). In the |
| 102 | // legacy parser, the T2529 hack is never applied if the template was |
| 103 | // already at the start of the line (the `!$piece['lineStart']` |
| 104 | // check in Parser::braceSubstitution where T2529 is handled), but |
| 105 | // that context (`$this->sol`) isn't passed through when Parsoid |
| 106 | // invokes the core preprocessor. Thus, when $this->sol is true, |
| 107 | // prepare to (if the following tokens warrant it) remove an unnecessary |
| 108 | // T2529 newline added by the legacy preprocessor. |
| 109 | $this->discardableNlTk = $token; |
| 110 | } |
| 111 | $this->tokenBuf[] = $token; |
| 112 | $this->sol = true; |
| 113 | return new TokenHandlerResult( [] ); |
| 114 | } |
| 115 | |
| 116 | /** |
| 117 | * @inheritDoc |
| 118 | */ |
| 119 | public function onEnd( EOFTk $token ): ?TokenHandlerResult { |
| 120 | $res = $this->onAny( $token ); |
| 121 | $this->reset(); |
| 122 | return $res; |
| 123 | } |
| 124 | |
| 125 | /** |
| 126 | * Clear start of line info |
| 127 | */ |
| 128 | private function clearSOL() { |
| 129 | // clear tsr and sol flag |
| 130 | $this->srcOffset = null; |
| 131 | $this->sol = false; |
| 132 | } |
| 133 | |
| 134 | /** |
| 135 | * Fully reprocess the output tokens from the tokenizer through |
| 136 | * all the other handlers in stage 2. |
| 137 | * |
| 138 | * @param int $srcOffset |
| 139 | * @param array $toks |
| 140 | * @param bool $popEOF |
| 141 | * @return array |
| 142 | */ |
| 143 | private function reprocessTokens( int $srcOffset, array $toks, bool $popEOF = false ): array { |
| 144 | // Update tsr |
| 145 | TokenUtils::shiftTokenTSR( $toks, $srcOffset ); |
| 146 | $pipe = $this->env->getPipelineFactory()->getPipeline( "tokens/x-mediawiki" ); |
| 147 | $pipe->init( [ |
| 148 | 'frame' => $this->manager->getFrame(), |
| 149 | 'toplevel' => $this->atTopLevel, |
| 150 | // The tokens should be reprocessed in the context of the original frame's source |
| 151 | 'srcText' => $this->manager->getFrame()->getSrcText() |
| 152 | ] ); |
| 153 | $toks = (array)$pipe->parse( $toks, [] ); |
| 154 | if ( $popEOF ) { |
| 155 | array_pop( $toks ); // pop EOFTk |
| 156 | } |
| 157 | return $toks; |
| 158 | } |
| 159 | |
| 160 | private function convertTokenToString( Token $token ): array { |
| 161 | $da = $token->dataParsoid; |
| 162 | $tsr = $da->tsr ?? null; |
| 163 | |
| 164 | if ( $tsr && $tsr->end > $tsr->start ) { |
| 165 | // > will only hold if these are valid numbers |
| 166 | $str = $tsr->substr( $this->manager->getFrame()->getSrcText() ); |
| 167 | // sol === false ensures that the pipe will not be parsed as a <td>/listItem again |
| 168 | $toks = $this->tokenizer->tokenizeSync( $str, [ 'sol' => false ] ); |
| 169 | return $this->reprocessTokens( $tsr->start, $toks, true ); |
| 170 | } elseif ( !empty( $da->autoInsertedStart ) && !empty( $da->autoInsertedEnd ) ) { |
| 171 | return [ '' ]; |
| 172 | } else { |
| 173 | // SSS FIXME: What about "!!" and "||"?? |
| 174 | switch ( $token->getName() ) { |
| 175 | case 'td': |
| 176 | return [ '|' ]; |
| 177 | case 'th': |
| 178 | return [ '!' ]; |
| 179 | case 'tr': |
| 180 | return [ '|-' ]; |
| 181 | case 'caption': |
| 182 | return [ $token instanceof TagTk ? '|+' : '' ]; |
| 183 | case 'table': |
| 184 | return [ $token instanceof EndTagTk ? '|}' : $token ]; |
| 185 | case 'listItem': |
| 186 | return [ implode( '', $token->getAttributeV( 'bullets' ) ) ]; |
| 187 | } |
| 188 | |
| 189 | // No conversion if we get here |
| 190 | return [ $token ]; |
| 191 | } |
| 192 | } |
| 193 | |
| 194 | /** |
| 195 | * @inheritDoc |
| 196 | */ |
| 197 | public function onAny( $token ): ?TokenHandlerResult { |
| 198 | try { |
| 199 | return $this->onAnyInternal( $token ); |
| 200 | } finally { |
| 201 | // Ensure we always clean up discardableNlTk and tplStartToken even |
| 202 | // in the presence of exceptions. |
| 203 | $this->discardableNlTk = null; |
| 204 | if ( $this->tplStartToken !== $token ) { |
| 205 | $this->tplStartToken = null; |
| 206 | } |
| 207 | } |
| 208 | } |
| 209 | |
| 210 | /** |
| 211 | * The legacy parser's "T2529 hack" attempts to ensure templates are |
| 212 | * always evaluated in start-of-line context by prepending a newline |
| 213 | * if necessary. However, it is inconsistent: in particular it |
| 214 | * only treats }| : ; # * as SOL-sensitive tokens, neglecting == |
| 215 | * (headings) and ! | |} (in table context). |
| 216 | * |
| 217 | * If we're using the core preprocessor for template expansion: |
| 218 | * - The core preprocessor as invoked by Parsoid will always insert the |
| 219 | * newline in the "T2529 cases" (even though it's not necessary; Parsoid |
| 220 | * is already in SOL mode) *HOWEVER* |
| 221 | * - As described in ::onNewline() above, the newline insertion is |
| 222 | * /supposed/ to be suppressed if the template was *already* |
| 223 | * at the start of the line. So we need to strip the unnecessarily |
| 224 | * added NlTk to avoid "extra" whitespace in Parsoid's expansion. |
| 225 | * Ex: "{{my-tpl}}" in sol-context which will get expanded to "\n*foo" |
| 226 | * but the "\n" wasn't necessary |
| 227 | * |
| 228 | * If we're in native preprocessor mode: |
| 229 | * - If we are in SOL state, we don't need to add a newline. |
| 230 | * - If we are not in SOL state, we need to insert a newline in 'T2529' cases. |
| 231 | * Ex: "{{my-tpl}}" in sol-context which expands to "*foo" but in |
| 232 | * non-sol context expands to "\n*foo" |
| 233 | * |
| 234 | * @param string $tokenName |
| 235 | */ |
| 236 | private function handleT2529Hack( string $tokenName ): void { |
| 237 | // Core's |
| 238 | if ( $tokenName === 'table' || $tokenName === 'listItem' ) { |
| 239 | // We're in a context when the core preprocessor would apply |
| 240 | // the "T2529 hack" to ensure start-of-line context. |
| 241 | if ( $this->discardableNlTk ) { |
| 242 | // We're using core preprocessor and were already at |
| 243 | // the start of the line, so the core preprocessor wouldn't |
| 244 | // actually have inserted a newline here. Swallow up ours. |
| 245 | array_pop( $this->tokenBuf ); |
| 246 | } elseif ( !$this->sol && |
| 247 | $this->tplStartToken && |
| 248 | $this->env->nativeTemplateExpansionEnabled() |
| 249 | ) { |
| 250 | // Native preprocessor; add a newline in "T2529 cases" |
| 251 | // for correct whitespace. (Remember that this only happens |
| 252 | // if we weren't already at the start of the line.) |
| 253 | // Add a newline & force SOL |
| 254 | $this->tokenBuf[] = new NlTk( null ); |
| 255 | $this->sol = true; |
| 256 | } |
| 257 | } |
| 258 | } |
| 259 | |
| 260 | /** |
| 261 | * @param mixed $token |
| 262 | * @return ?TokenHandlerResult |
| 263 | */ |
| 264 | public function onAnyInternal( $token ): ?TokenHandlerResult { |
| 265 | $self = $this; |
| 266 | $this->env->log( 'trace/tsp', $this->pipelineId, |
| 267 | static function () use ( $self, $token ) { |
| 268 | return "(indep=" . ( $self->inIndependentParse ? "yes" : "no " ) . |
| 269 | ";sol=" . ( $self->sol ? "yes" : "no " ) . |
| 270 | PHPUtils::jsonEncode( $token ); |
| 271 | } |
| 272 | ); |
| 273 | |
| 274 | $tokens = [ $token ]; |
| 275 | $tc = TokenUtils::getTokenType( $token ); |
| 276 | switch ( $tc ) { |
| 277 | case 'string': |
| 278 | // While we are buffering newlines to suppress them |
| 279 | // in case we see a category, buffer all intervening |
| 280 | // white-space as well. |
| 281 | if ( count( $this->tokenBuf ) > 0 && preg_match( '/^\s*$/D', $token ) ) { |
| 282 | $this->tokenBuf[] = $token; |
| 283 | return new TokenHandlerResult( [] ); |
| 284 | } |
| 285 | |
| 286 | // This is only applicable where we use Parsoid's (broken) native preprocessor. |
| 287 | // This supports scenarios like "{{1x|*bar}}". When "{{{1}}}" is tokenized |
| 288 | // "*bar" isn't available and so won't become a list. |
| 289 | // FIXME: {{1x|1===foo==}} will still be broken. So, this fix below is somewhat |
| 290 | // independent of T2529 for our broken preprocessor but we are restricting the |
| 291 | // fix to T2529. |
| 292 | $T2529hack = false; |
| 293 | if ( $this->env->nativeTemplateExpansionEnabled() && |
| 294 | $this->tplStartToken && |
| 295 | preg_match( '/^(?:{\\||[:;#*])/', $token ) |
| 296 | ) { |
| 297 | // Add a newline & force SOL |
| 298 | $T2529hack = true; |
| 299 | // Remove newline insertion in the core preprocessor |
| 300 | // only occurs if we weren't already at the start of |
| 301 | // the line (see discussion in ::onNewline() above). |
| 302 | if ( !$this->sol ) { |
| 303 | $this->tokenBuf[] = new NlTk( null ); |
| 304 | $this->sol = true; |
| 305 | } |
| 306 | } |
| 307 | |
| 308 | if ( $this->sol ) { |
| 309 | // Attempt to match "{|" after a newline and convert |
| 310 | // it to a table token. |
| 311 | if ( $this->inIndependentParse && str_starts_with( $token, '{|' ) ) { |
| 312 | // Reparse string with the 'table_start_tag' rule |
| 313 | // and fully reprocess them. |
| 314 | $retoks = $this->tokenizer->tokenizeAs( $token, 'table_start_tag', /* sol */true ); |
| 315 | if ( $retoks === false ) { |
| 316 | // XXX: The string begins with table start syntax, |
| 317 | // we really shouldn't be here. Anything else on the |
| 318 | // line would get swallowed up as attributes. |
| 319 | $this->env->log( 'error', 'Failed to tokenize table start tag.' ); |
| 320 | $this->clearSOL(); |
| 321 | } else { |
| 322 | $tokens = $this->reprocessTokens( $this->srcOffset, $retoks ); |
| 323 | $this->wikiTableNesting++; |
| 324 | $this->lastConvertedTableCellToken = null; |
| 325 | } |
| 326 | } elseif ( $this->inIndependentParse && $T2529hack ) { // {| has been handled above |
| 327 | $retoks = $this->tokenizer->tokenizeAs( $token, 'list_item', /* sol */true ); |
| 328 | if ( $retoks === false ) { |
| 329 | $this->env->log( 'error', 'Failed to tokenize list item.' ); |
| 330 | $this->clearSOL(); |
| 331 | } else { |
| 332 | $tokens = $this->reprocessTokens( $this->srcOffset, $retoks ); |
| 333 | } |
| 334 | } elseif ( preg_match( '/^\s*$/D', $token ) ) { |
| 335 | // White-space doesn't change SOL state |
| 336 | // Update srcOffset |
| 337 | $this->srcOffset += strlen( $token ); |
| 338 | } else { |
| 339 | $this->clearSOL(); |
| 340 | } |
| 341 | } else { |
| 342 | $this->clearSOL(); |
| 343 | } |
| 344 | break; |
| 345 | |
| 346 | case 'CommentTk': |
| 347 | // Comments don't change SOL state |
| 348 | // Update srcOffset |
| 349 | $this->srcOffset = $token->dataParsoid->tsr->end ?? null; |
| 350 | break; |
| 351 | |
| 352 | case 'SelfclosingTagTk': |
| 353 | if ( $token->getName() === 'meta' && ( $token->dataParsoid->stx ?? '' ) !== 'html' ) { |
| 354 | if ( TokenUtils::hasTypeOf( $token, 'mw:Transclusion' ) ) { |
| 355 | $this->tplStartToken = $token; |
| 356 | } |
| 357 | $this->srcOffset = $token->dataParsoid->tsr->end ?? null; |
| 358 | if ( count( $this->tokenBuf ) > 0 && |
| 359 | TokenUtils::hasTypeOf( $token, 'mw:Transclusion' ) |
| 360 | ) { |
| 361 | // If we have buffered newlines, we might very well encounter |
| 362 | // a category link, so continue buffering. |
| 363 | $this->tokenBuf[] = $token; |
| 364 | return new TokenHandlerResult( [] ); |
| 365 | } |
| 366 | } elseif ( $token->getName() === 'link' && |
| 367 | $token->getAttributeV( 'rel' ) === 'mw:PageProp/Category' |
| 368 | ) { |
| 369 | // Replace buffered newline & whitespace tokens with mw:EmptyLine |
| 370 | // meta-tokens. This tunnels them through the rest of the transformations |
| 371 | // without affecting them. During HTML building, they are expanded |
| 372 | // back to newlines / whitespace. |
| 373 | $n = count( $this->tokenBuf ); |
| 374 | if ( $n > 0 ) { |
| 375 | $i = 0; |
| 376 | while ( $i < $n && |
| 377 | !( $this->tokenBuf[$i] instanceof SelfclosingTagTk ) |
| 378 | ) { |
| 379 | $i++; |
| 380 | } |
| 381 | |
| 382 | $dp = new DataParsoid; |
| 383 | $dp->tokens = array_slice( $this->tokenBuf, 0, $i ); |
| 384 | $toks = [ |
| 385 | new SelfclosingTagTk( 'meta', |
| 386 | [ new KV( 'typeof', 'mw:EmptyLine' ) ], |
| 387 | $dp |
| 388 | ) |
| 389 | ]; |
| 390 | if ( $i < $n ) { |
| 391 | $toks[] = $this->tokenBuf[$i]; |
| 392 | if ( $i + 1 < $n ) { |
| 393 | $dp = new DataParsoid; |
| 394 | $dp->tokens = array_slice( $this->tokenBuf, $i + 1 ); |
| 395 | $toks[] = new SelfclosingTagTk( 'meta', |
| 396 | [ new KV( 'typeof', 'mw:EmptyLine' ) ], |
| 397 | $dp |
| 398 | ); |
| 399 | } |
| 400 | } |
| 401 | $tokens = array_merge( $toks, $tokens ); |
| 402 | $this->tokenBuf = []; |
| 403 | } |
| 404 | $this->clearSOL(); |
| 405 | } else { |
| 406 | $this->clearSOL(); |
| 407 | } |
| 408 | break; |
| 409 | |
| 410 | case 'TagTk': |
| 411 | if ( $this->inIndependentParse && !TokenUtils::isHTMLTag( $token ) ) { |
| 412 | $tokenName = $token->getName(); |
| 413 | $this->handleT2529Hack( $tokenName ); |
| 414 | if ( $tokenName === 'listItem' && isset( $this->options['attrExpansion'] ) ) { |
| 415 | // Convert list items back to bullet wikitext in attribute context |
| 416 | $tokens = $this->convertTokenToString( $token ); |
| 417 | } elseif ( $tokenName === 'table' ) { |
| 418 | $this->lastConvertedTableCellToken = null; |
| 419 | $this->wikiTableNesting++; |
| 420 | } elseif ( in_array( $tokenName, [ 'td', 'th', 'tr', 'caption' ], true ) ) { |
| 421 | if ( $this->wikiTableNesting === 0 ) { |
| 422 | if ( $token->getName() === 'td' || $token->getName() === 'th' ) { |
| 423 | $this->lastConvertedTableCellToken = $token; |
| 424 | } |
| 425 | $tokens = $this->convertTokenToString( $token ); |
| 426 | } else { |
| 427 | $this->lastConvertedTableCellToken = null; |
| 428 | } |
| 429 | } |
| 430 | } |
| 431 | $this->clearSOL(); |
| 432 | break; |
| 433 | |
| 434 | case 'EndTagTk': |
| 435 | if ( $this->inIndependentParse && !TokenUtils::isHTMLTag( $token ) ) { |
| 436 | if ( $this->wikiTableNesting > 0 ) { |
| 437 | if ( $token->getName() === 'table' ) { |
| 438 | $this->lastConvertedTableCellToken = null; |
| 439 | $this->wikiTableNesting--; |
| 440 | } |
| 441 | } elseif ( $token->getName() === 'table' || $token->getName() === 'caption' ) { |
| 442 | // Convert this to "|}" |
| 443 | $tokens = $this->convertTokenToString( $token ); |
| 444 | } |
| 445 | } |
| 446 | $this->clearSOL(); |
| 447 | break; |
| 448 | |
| 449 | default: |
| 450 | break; |
| 451 | } |
| 452 | |
| 453 | // Emit buffered newlines (and a transclusion meta-token, if any) |
| 454 | if ( count( $this->tokenBuf ) > 0 ) { |
| 455 | $tokens = array_merge( $this->tokenBuf, $tokens ); |
| 456 | $this->tokenBuf = []; |
| 457 | } |
| 458 | return new TokenHandlerResult( $tokens ); |
| 459 | } |
| 460 | } |