Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
10.14% |
14 / 138 |
|
0.00% |
0 / 15 |
CRAP | |
0.00% |
0 / 1 |
| PegTokenizer | |
10.14% |
14 / 138 |
|
0.00% |
0 / 15 |
1581.12 | |
0.00% |
0 / 1 |
| __construct | |
0.00% |
0 / 9 |
|
0.00% |
0 / 1 |
6 | |||
| initGrammar | |
0.00% |
0 / 2 |
|
0.00% |
0 / 1 |
12 | |||
| getOptions | |
0.00% |
0 / 1 |
|
0.00% |
0 / 1 |
2 | |||
| getFrame | |
0.00% |
0 / 1 |
|
0.00% |
0 / 1 |
2 | |||
| process | |
0.00% |
0 / 6 |
|
0.00% |
0 / 1 |
6 | |||
| processChunkily | |
0.00% |
0 / 21 |
|
0.00% |
0 / 1 |
20 | |||
| onlyIncludeOffsets | |
0.00% |
0 / 11 |
|
0.00% |
0 / 1 |
30 | |||
| finalize | |
0.00% |
0 / 1 |
|
0.00% |
0 / 1 |
2 | |||
| tokenizeSync | |
0.00% |
0 / 57 |
|
0.00% |
0 / 1 |
240 | |||
| tokenizeAs | |
81.82% |
9 / 11 |
|
0.00% |
0 / 1 |
2.02 | |||
| tokenizeURL | |
0.00% |
0 / 1 |
|
0.00% |
0 / 1 |
2 | |||
| tokenizeTemplate | |
0.00% |
0 / 6 |
|
0.00% |
0 / 1 |
12 | |||
| tokenizeTemplate3 | |
83.33% |
5 / 6 |
|
0.00% |
0 / 1 |
3.04 | |||
| tokenizeTableCellAttributes | |
0.00% |
0 / 1 |
|
0.00% |
0 / 1 |
2 | |||
| resetState | |
0.00% |
0 / 4 |
|
0.00% |
0 / 1 |
6 | |||
| 1 | <?php |
| 2 | declare( strict_types = 1 ); |
| 3 | |
| 4 | namespace Wikimedia\Parsoid\Wt2Html; |
| 5 | |
| 6 | use Generator; |
| 7 | use Wikimedia\Assert\Assert; |
| 8 | use Wikimedia\Parsoid\Config\Env; |
| 9 | use Wikimedia\Parsoid\Core\Source; |
| 10 | use Wikimedia\Parsoid\Core\SourceRange; |
| 11 | use Wikimedia\Parsoid\Core\SourceString; |
| 12 | use Wikimedia\Parsoid\DOM\DocumentFragment; |
| 13 | use Wikimedia\Parsoid\DOM\Element; |
| 14 | use Wikimedia\Parsoid\Tokens\EOFTk; |
| 15 | use Wikimedia\Parsoid\Tokens\SelfclosingTagTk; |
| 16 | use Wikimedia\Parsoid\Tokens\Token; |
| 17 | use Wikimedia\Parsoid\Utils\PHPUtils; |
| 18 | use Wikimedia\Parsoid\Utils\TokenUtils; |
| 19 | use Wikimedia\WikiPEG\SyntaxError; |
| 20 | |
| 21 | /** |
| 22 | * Tokenizer for wikitext, using WikiPEG and a |
| 23 | * separate PEG grammar file (Grammar.pegphp) |
| 24 | */ |
| 25 | class PegTokenizer extends PipelineStage { |
| 26 | private array $options; |
| 27 | /** @var Grammar|TracingGrammar|null */ |
| 28 | private $grammar = null; |
| 29 | private bool $tracing; |
| 30 | /** |
| 31 | * No need to retokenize identical strings |
| 32 | * Cache <src,startRule> --> token array. |
| 33 | * Expected benefits: |
| 34 | * - same expanded template source used multiple times on a page |
| 35 | * - convertToString calls |
| 36 | * - calls from TableFixups and elsewhere to tokenize* methods |
| 37 | */ |
| 38 | private PipelineContentCache $cache; |
| 39 | |
| 40 | public function __construct( Env $env, array $options = [], string $stageId = "" ) { |
| 41 | parent::__construct( $env ); |
| 42 | $this->env = $env; |
| 43 | $this->options = $options; |
| 44 | $this->tracing = $env->hasTraceFlag( 'grammar' ) || |
| 45 | // Allow substitution of a custom tracer for unit testing |
| 46 | isset( $options['tracer'] ); |
| 47 | // Cache only on seeing the same source the second time. |
| 48 | // This minimizes cache bloat & token cloning penalties. |
| 49 | $this->cache = $this->env->getCache( |
| 50 | "PegTokenizer", |
| 51 | [ "repeatThreshold" => 1, "cloneValue" => true ] |
| 52 | ); |
| 53 | } |
| 54 | |
| 55 | private function initGrammar(): void { |
| 56 | if ( !$this->grammar ) { |
| 57 | $this->grammar = $this->tracing ? new TracingGrammar : new Grammar; |
| 58 | } |
| 59 | } |
| 60 | |
| 61 | /** |
| 62 | * Get the constructor options. |
| 63 | * |
| 64 | * @internal |
| 65 | * @return array |
| 66 | */ |
| 67 | public function getOptions(): array { |
| 68 | return $this->options; |
| 69 | } |
| 70 | |
| 71 | public function getFrame(): Frame { |
| 72 | return $this->frame ?? $this->env->topFrame; |
| 73 | } |
| 74 | |
| 75 | /** |
| 76 | * See PipelineStage::process docs as well. This doc block refines |
| 77 | * the generic arg types to be specific to this pipeline stage. |
| 78 | * |
| 79 | * @param string|array|DocumentFragment|Element $input |
| 80 | * Wikitext to tokenize. In practice this should be a string. |
| 81 | * @param array{sol:bool} $options |
| 82 | * - atTopLevel: (bool) Whether we are processing the top-level document |
| 83 | * - sol: (bool) Whether input should be processed in start-of-line context |
| 84 | * |
| 85 | * @return array The token array |
| 86 | * @throws SyntaxError |
| 87 | */ |
| 88 | public function process( |
| 89 | string|array|DocumentFragment|Element $input, |
| 90 | array $options |
| 91 | ): array|Element|DocumentFragment { |
| 92 | Assert::invariant( is_string( $input ), "Input should be a string" ); |
| 93 | $result = $this->tokenizeSync( $input, $options, $exception ); |
| 94 | if ( $result === false ) { |
| 95 | // Should never happen. |
| 96 | throw $exception; |
| 97 | } |
| 98 | |
| 99 | // Downstream stages expect EOFTk at the end of the token stream. |
| 100 | $result[] = new EOFTk(); |
| 101 | return $result; |
| 102 | } |
| 103 | |
| 104 | /** |
| 105 | * The text is tokenized in chunks (one per top-level block). |
| 106 | * |
| 107 | * @param string|array|DocumentFragment|Element $input |
| 108 | * Wikitext to tokenize. In practice this should be a string. |
| 109 | * @param array{atTopLevel:bool,sol:bool} $options |
| 110 | * - atTopLevel: (bool) Whether we are processing the top-level document |
| 111 | * - sol (bool) Whether text should be processed in start-of-line context. |
| 112 | * @return Generator<list<Token|string>> |
| 113 | */ |
| 114 | public function processChunkily( |
| 115 | string|array|DocumentFragment|Element $input, |
| 116 | array $options |
| 117 | ): Generator { |
| 118 | if ( !$this->grammar ) { |
| 119 | $this->initGrammar(); |
| 120 | } |
| 121 | |
| 122 | Assert::invariant( is_string( $input ), "Input should be a string" ); |
| 123 | Assert::invariant( isset( $options['sol'] ), "Sol should be set" ); |
| 124 | |
| 125 | // Kick it off! |
| 126 | $args = [ |
| 127 | 'env' => $this->env, |
| 128 | 'pipelineId' => $this->getPipelineId(), |
| 129 | 'pegTokenizer' => $this, |
| 130 | 'pipelineOffset' => $this->srcOffsets->start ?? 0, |
| 131 | 'source' => $this->srcOffsets->source, |
| 132 | 'sol' => $options['sol'], |
| 133 | 'stream' => true, |
| 134 | 'startRule' => 'start_async', |
| 135 | ]; |
| 136 | |
| 137 | if ( $this->tracing ) { |
| 138 | $args['tracer'] = $this->options['tracer'] ?? new Tracer( $input ); |
| 139 | } |
| 140 | |
| 141 | foreach ( $this->onlyIncludeOffsets( $input, $args ) as [ $start, $end ] ) { |
| 142 | $piece = substr( $input, $start, $end - $start ); |
| 143 | // Wrap wikipeg's generator with our own generator |
| 144 | // to track time usage. |
| 145 | // @phan-suppress-next-line PhanTypeInvalidYieldFrom |
| 146 | yield from $this->grammar->parse( $piece, [ |
| 147 | 'pipelineOffset' => $args['pipelineOffset'] + $start, |
| 148 | ] + $args ); |
| 149 | } |
| 150 | } |
| 151 | |
| 152 | /** |
| 153 | * Provide the offsets into $input needed for <onlyinclude> processing |
| 154 | * if `inTemplate` mode. Otherwise just return the start and end |
| 155 | * of the string. |
| 156 | */ |
| 157 | private function onlyIncludeOffsets( string $input, array $args ): array { |
| 158 | // Handle <onlyinclude> |
| 159 | if ( |
| 160 | ( $this->options['inTemplate'] ?? false ) && |
| 161 | str_contains( $input, '<onlyinclude>' ) && |
| 162 | str_contains( $input, '</onlyinclude>' ) |
| 163 | ) { |
| 164 | try { |
| 165 | return $this->grammar->parse( $input, [ |
| 166 | 'stream' => false, |
| 167 | 'startRule' => 'preproc_find_only_include', |
| 168 | 'pipelineOffset' => 0, |
| 169 | ] + $args ); |
| 170 | } catch ( SyntaxError ) { |
| 171 | /* ignore, fall through to process the whole input */ |
| 172 | $this->env->log( 'warn', "Couldn't extract <onlyinclude>" ); |
| 173 | } |
| 174 | } |
| 175 | return [ [ 0, strlen( $input ) ] ]; |
| 176 | } |
| 177 | |
| 178 | /** |
| 179 | * @inheritDoc |
| 180 | */ |
| 181 | public function finalize(): Generator { |
| 182 | yield [ new EOFTk() ]; |
| 183 | } |
| 184 | |
| 185 | /** |
| 186 | * Tokenize via a rule passed in as an arg. |
| 187 | * The text is tokenized synchronously in one shot. |
| 188 | * |
| 189 | * @param string $text |
| 190 | * @param array{sol:bool} $args |
| 191 | * - sol: (bool) Whether input should be processed in start-of-line context. |
| 192 | * - startRule: (string) which tokenizer rule to tokenize with |
| 193 | * @param SyntaxError|null &$exception a syntax error, if thrown. |
| 194 | * @return array|false The token array, or false for a syntax error |
| 195 | */ |
| 196 | public function tokenizeSync( string $text, array $args, &$exception = null ) { |
| 197 | if ( !$this->grammar ) { |
| 198 | $this->initGrammar(); |
| 199 | } |
| 200 | Assert::invariant( isset( $args['sol'] ), "Sol should be set" ); |
| 201 | $args += [ |
| 202 | 'pegTokenizer' => $this, |
| 203 | 'pipelineId' => $this->getPipelineId(), |
| 204 | 'pipelineOffset' => $this->srcOffsets->start ?? 0, |
| 205 | 'source' => $this->srcOffsets->source ?? null, |
| 206 | 'startRule' => 'start', |
| 207 | 'env' => $this->env |
| 208 | ]; |
| 209 | Assert::invariant( $args['startRule'] !== null, 'null start rule' ); |
| 210 | Assert::invariant( !( $args['stream'] ?? false ), 'synchronous parse' ); |
| 211 | |
| 212 | // crc32 is much faster than md5 and since we are verifying a |
| 213 | // $text match when reusing cache contents, hash collisions are okay. |
| 214 | // |
| 215 | // NOTE about inclusion of pipelineOffset in the cache key: |
| 216 | // The PEG tokenizer returns tokens with offsets shifted by |
| 217 | // $args['pipelineOffset'], so we cannot reuse tokens across |
| 218 | // differing values of this option. If required, we could refactor |
| 219 | // to move that and the logging code into this file. |
| 220 | $cacheKey = crc32( $text ) . |
| 221 | "|" . (int)$args['sol'] . |
| 222 | "|" . $args['startRule'] . |
| 223 | "|" . $args['pipelineOffset'] . |
| 224 | "|" . ( $args['source'] ? spl_object_id( $args['source'] ) : "" ) . |
| 225 | "|" . ( $this->options['enableOnlyInclude'] ?? false ); |
| 226 | $res = $this->cache->lookup( $cacheKey, $text ); |
| 227 | if ( $res !== null ) { |
| 228 | $tokens = $res['value']; |
| 229 | // NOTE: Since $args['source'] is part of the cache key, |
| 230 | // we don't need to reset source in SourceRange properties in $tokens. |
| 231 | return $tokens; |
| 232 | } |
| 233 | |
| 234 | if ( $this->tracing ) { |
| 235 | $args['tracer'] = $this->options['tracer'] ?? new Tracer( $text ); |
| 236 | } |
| 237 | |
| 238 | $start = null; |
| 239 | $profile = null; |
| 240 | if ( $this->env->profiling() ) { |
| 241 | $profile = $this->env->getCurrentProfile(); |
| 242 | $start = hrtime( true ); |
| 243 | } |
| 244 | |
| 245 | try { |
| 246 | if ( $this->options['enableOnlyInclude'] ?? false ) { |
| 247 | $toks = []; |
| 248 | foreach ( $this->onlyIncludeOffsets( $text, $args ) as [ $start, $end ] ) { |
| 249 | $piece = substr( $text, $start, $end - $start ); |
| 250 | $result = $this->grammar->parse( $piece, [ |
| 251 | 'pipelineOffset' => $args['pipelineOffset'] + $start, |
| 252 | ] + $args ); |
| 253 | if ( !is_array( $result ) ) { |
| 254 | $result = [ $result ]; |
| 255 | } |
| 256 | // The 'start' and 'start_async' rules manually call |
| 257 | // ::shiftTokenTSR before returning tokens. For all |
| 258 | // others, we still need to perform the shift by the |
| 259 | // requested $args['pipelineOffset']. |
| 260 | if ( $args['startRule'] !== 'start' ) { |
| 261 | TokenUtils::shiftTokenTSR( |
| 262 | $result, $args['pipelineOffset'] + $start |
| 263 | ); |
| 264 | } |
| 265 | PHPUtils::pushArray( $toks, $result ); |
| 266 | } |
| 267 | } else { |
| 268 | $toks = $this->grammar->parse( $text, $args ); |
| 269 | // The 'start' and 'start_async' rules manually call |
| 270 | // ::shiftTokenTSR before returning tokens. For all |
| 271 | // others, we still need to perform the shift by the |
| 272 | // requested $args['pipelineOffset']. |
| 273 | if ( $args['startRule'] !== 'start' ) { |
| 274 | TokenUtils::shiftTokenTSR( |
| 275 | is_array( $toks ) ? $toks : [ $toks ], $args['pipelineOffset'] |
| 276 | ); |
| 277 | } |
| 278 | } |
| 279 | } catch ( SyntaxError $e ) { |
| 280 | $exception = $e; |
| 281 | return false; |
| 282 | } |
| 283 | |
| 284 | if ( $profile ) { |
| 285 | $profile->bumpTimeUse( 'PEG', hrtime( true ) - $start, 'PEG' ); |
| 286 | } |
| 287 | |
| 288 | if ( is_array( $toks ) ) { |
| 289 | $this->cache->cache( $cacheKey, $toks, $this->frame?->getSource(), $text ); |
| 290 | } |
| 291 | |
| 292 | return $toks; |
| 293 | } |
| 294 | |
| 295 | /** |
| 296 | * Tokenizes a string as a rule |
| 297 | * |
| 298 | * @param string|Source $text The input text |
| 299 | * @param string $rule The rule name |
| 300 | * @param bool $sol Start of line flag |
| 301 | * @param ?SourceRange $tsr The optional position of $text in the Source |
| 302 | * @return array|false Array of tokens/strings or false on error |
| 303 | */ |
| 304 | public function tokenizeAs( string|Source $text, string $rule, bool $sol, ?SourceRange $tsr = null ) { |
| 305 | if ( $text instanceof Source ) { |
| 306 | $source = $text; |
| 307 | $text = $source->getSrcText(); |
| 308 | } else { |
| 309 | $source = new SourceString( $text ); |
| 310 | } |
| 311 | $args = [ |
| 312 | 'startRule' => $rule, |
| 313 | 'sol' => $sol, |
| 314 | 'pipelineOffset' => $tsr->start ?? 0, |
| 315 | 'source' => $tsr->source ?? $source, |
| 316 | ]; |
| 317 | return $this->tokenizeSync( $text, $args ); |
| 318 | } |
| 319 | |
| 320 | /** |
| 321 | * Tokenize a URL. |
| 322 | * @param string $text |
| 323 | * @return array|false Array of tokens/strings or false on error |
| 324 | */ |
| 325 | public function tokenizeURL( string $text ) { |
| 326 | // XXX T405759 This returns tokens with a unique (not top-level) |
| 327 | // source in the TSR; if this is retokenizing part of the top-level |
| 328 | // source this should pass srcOffsets. |
| 329 | return $this->tokenizeAs( $text, 'url', /* sol */true ); |
| 330 | } |
| 331 | |
| 332 | /** |
| 333 | * (Re)tokenize a template / parser function. |
| 334 | * @param string $text |
| 335 | * @param SourceRange $tsr The location of $text in the Source |
| 336 | * @return SelfclosingTagTk|false A template token or false on error |
| 337 | */ |
| 338 | public function tokenizeTemplate( string $text, SourceRange $tsr ) { |
| 339 | $toks = $this->tokenizeAs( $text, 'template', false, $tsr ); |
| 340 | if ( !is_array( $toks ) ) { |
| 341 | $toks = [ $toks ]; |
| 342 | } |
| 343 | if ( count( $toks ) === 1 ) { |
| 344 | return $toks[0]; |
| 345 | } |
| 346 | return false; |
| 347 | } |
| 348 | |
| 349 | /** |
| 350 | * (Re)tokenize a v3 template / parser function. |
| 351 | * @param string $text |
| 352 | * @param SourceRange $tsr The location of $text in the Source |
| 353 | * @return SelfclosingTagTk|false A template3 token or false on error |
| 354 | */ |
| 355 | public function tokenizeTemplate3( string $text, SourceRange $tsr ) { |
| 356 | $toks = $this->tokenizeAs( $text, 'template3', false, $tsr ); |
| 357 | if ( !is_array( $toks ) ) { |
| 358 | $toks = [ $toks ]; |
| 359 | } |
| 360 | if ( count( $toks ) === 1 ) { |
| 361 | return $toks[0]; |
| 362 | } |
| 363 | return false; |
| 364 | } |
| 365 | |
| 366 | /** |
| 367 | * Tokenize table cell attributes. |
| 368 | * @param string $text |
| 369 | * @param bool $sol |
| 370 | * @return array|false Array of tokens/strings or false on error |
| 371 | */ |
| 372 | public function tokenizeTableCellAttributes( string $text, bool $sol ) { |
| 373 | // XXX T405759 This returns tokens with a unique (not top-level) |
| 374 | // source in the TSR; if this is retokenizing part of the top-level |
| 375 | // source this should pass srcOffsets. |
| 376 | return $this->tokenizeAs( $text, 'row_syntax_table_args', $sol ); |
| 377 | } |
| 378 | |
| 379 | /** |
| 380 | * @inheritDoc |
| 381 | */ |
| 382 | public function resetState( array $options ): void { |
| 383 | TokenizerUtils::resetAnnotationIncludeRegex(); |
| 384 | if ( $this->grammar ) { |
| 385 | $this->grammar->resetState(); |
| 386 | } |
| 387 | parent::resetState( $options ); |
| 388 | } |
| 389 | } |