Code Coverage |
||||||||||
Lines |
Functions and Methods |
Classes and Traits |
||||||||
| Total | |
0.00% |
0 / 581 |
|
0.00% |
0 / 22 |
CRAP | |
0.00% |
0 / 1 |
| WikitextEscapeHandlers | |
0.00% |
0 / 581 |
|
0.00% |
0 / 22 |
61256 | |
0.00% |
0 / 1 |
| __construct | |
0.00% |
0 / 3 |
|
0.00% |
0 / 1 |
2 | |||
| startsOnANewLine | |
0.00% |
0 / 3 |
|
0.00% |
0 / 1 |
6 | |||
| hasBlocksOnLine | |
0.00% |
0 / 17 |
|
0.00% |
0 / 1 |
110 | |||
| hasLeadingEscapableQuoteChar | |
0.00% |
0 / 9 |
|
0.00% |
0 / 1 |
20 | |||
| hasTrailingEscapableQuoteChar | |
0.00% |
0 / 9 |
|
0.00% |
0 / 1 |
20 | |||
| escapedIBSiblingNodeText | |
0.00% |
0 / 13 |
|
0.00% |
0 / 1 |
30 | |||
| isFirstContentNode | |
0.00% |
0 / 1 |
|
0.00% |
0 / 1 |
2 | |||
| liHandler | |
0.00% |
0 / 9 |
|
0.00% |
0 / 1 |
42 | |||
| thHandler | |
0.00% |
0 / 2 |
|
0.00% |
0 / 1 |
6 | |||
| mediaOptionHandler | |
0.00% |
0 / 1 |
|
0.00% |
0 / 1 |
6 | |||
| wikilinkHandler | |
0.00% |
0 / 1 |
|
0.00% |
0 / 1 |
2 | |||
| aHandler | |
0.00% |
0 / 1 |
|
0.00% |
0 / 1 |
2 | |||
| tdHandler | |
0.00% |
0 / 15 |
|
0.00% |
0 / 1 |
182 | |||
| tokenizeStr | |
0.00% |
0 / 6 |
|
0.00% |
0 / 1 |
2 | |||
| textCanParseAsLink | |
0.00% |
0 / 52 |
|
0.00% |
0 / 1 |
552 | |||
| hasWikitextTokens | |
0.00% |
0 / 68 |
|
0.00% |
0 / 1 |
1640 | |||
| nowikiWrap | |
0.00% |
0 / 8 |
|
0.00% |
0 / 1 |
12 | |||
| escapedText | |
0.00% |
0 / 78 |
|
0.00% |
0 / 1 |
600 | |||
| escapeWikitext | |
0.00% |
0 / 111 |
|
0.00% |
0 / 1 |
2652 | |||
| appendStr | |
0.00% |
0 / 37 |
|
0.00% |
0 / 1 |
552 | |||
| escapeTplArgWT | |
0.00% |
0 / 127 |
|
0.00% |
0 / 1 |
812 | |||
| escapeLinkContent | |
0.00% |
0 / 10 |
|
0.00% |
0 / 1 |
6 | |||
| 1 | <?php |
| 2 | declare( strict_types = 1 ); |
| 3 | |
| 4 | namespace Wikimedia\Parsoid\Html2Wt; |
| 5 | |
| 6 | use Wikimedia\Assert\Assert; |
| 7 | use Wikimedia\Parsoid\Config\Env; |
| 8 | use Wikimedia\Parsoid\DOM\Element; |
| 9 | use Wikimedia\Parsoid\DOM\Node; |
| 10 | use Wikimedia\Parsoid\Tokens\EndTagTk; |
| 11 | use Wikimedia\Parsoid\Tokens\EOFTk; |
| 12 | use Wikimedia\Parsoid\Tokens\SourceRange; |
| 13 | use Wikimedia\Parsoid\Tokens\TagTk; |
| 14 | use Wikimedia\Parsoid\Tokens\Token; |
| 15 | use Wikimedia\Parsoid\Utils\DiffDOMUtils; |
| 16 | use Wikimedia\Parsoid\Utils\DOMCompat; |
| 17 | use Wikimedia\Parsoid\Utils\DOMUtils; |
| 18 | use Wikimedia\Parsoid\Utils\PHPUtils; |
| 19 | use Wikimedia\Parsoid\Utils\TokenUtils; |
| 20 | use Wikimedia\Parsoid\Utils\Utils; |
| 21 | use Wikimedia\Parsoid\Utils\WTUtils; |
| 22 | use Wikimedia\Parsoid\Wikitext\Consts; |
| 23 | use Wikimedia\Parsoid\Wt2Html\PegTokenizer; |
| 24 | |
| 25 | class WikitextEscapeHandlers { |
| 26 | |
| 27 | private const LINKS_ESCAPE_RE = '/(\[\[)|(\]\])|(-\{)|(^[^\[]*\]$)/D'; |
| 28 | |
| 29 | /** @var Env */ |
| 30 | private $env; |
| 31 | |
| 32 | /** @var ?string */ |
| 33 | private $extName; |
| 34 | |
| 35 | /** |
| 36 | * @var PegTokenizer |
| 37 | */ |
| 38 | private $tokenizer; |
| 39 | |
| 40 | public function __construct( Env $env, ?string $extName ) { |
| 41 | $this->env = $env; |
| 42 | $this->extName = $extName; |
| 43 | $this->tokenizer = new PegTokenizer( $env ); |
| 44 | } |
| 45 | |
| 46 | /** |
| 47 | * Ignore the cases where the serializer adds newlines not present in the dom |
| 48 | * @param Node $node |
| 49 | * @return bool |
| 50 | */ |
| 51 | private static function startsOnANewLine( Node $node ): bool { |
| 52 | $name = DOMCompat::nodeName( $node ); |
| 53 | return TokenUtils::tagOpensBlockScope( $name ) && |
| 54 | !WTUtils::isLiteralHTMLNode( $node ); |
| 55 | } |
| 56 | |
| 57 | /** |
| 58 | * Look ahead on current line for block content |
| 59 | * |
| 60 | * @param Node $node |
| 61 | * @param bool $first |
| 62 | * @return bool |
| 63 | */ |
| 64 | private static function hasBlocksOnLine( Node $node, bool $first ): bool { |
| 65 | // special case for firstNode: |
| 66 | // we're at sol so ignore possible \n at first char |
| 67 | if ( $first ) { |
| 68 | $textContent = $node->textContent; |
| 69 | $offset = strlen( $textContent ) ? 1 : 0; |
| 70 | if ( strpos( $textContent, "\n", $offset ) !== false ) { |
| 71 | return false; |
| 72 | } |
| 73 | $node = $node->nextSibling; |
| 74 | } |
| 75 | |
| 76 | while ( $node ) { |
| 77 | if ( $node instanceof Element ) { |
| 78 | if ( DOMUtils::isWikitextBlockNode( $node ) ) { |
| 79 | return !self::startsOnANewLine( $node ); |
| 80 | } |
| 81 | if ( $node->hasChildNodes() ) { |
| 82 | if ( self::hasBlocksOnLine( $node->firstChild, false ) ) { |
| 83 | return true; |
| 84 | } |
| 85 | } |
| 86 | } else { |
| 87 | if ( str_contains( $node->textContent, "\n" ) ) { |
| 88 | return false; |
| 89 | } |
| 90 | } |
| 91 | $node = $node->nextSibling; |
| 92 | } |
| 93 | return false; |
| 94 | } |
| 95 | |
| 96 | /** |
| 97 | * @param string $text |
| 98 | * @param array $opts [ 'node' => Node ] |
| 99 | * @return bool |
| 100 | */ |
| 101 | private static function hasLeadingEscapableQuoteChar( string $text, array $opts ): bool { |
| 102 | /** @var Node $node */ |
| 103 | $node = $opts['node']; |
| 104 | // Use 'node.textContent' to do the tests since it hasn't had newlines |
| 105 | // stripped out from it. |
| 106 | // Ex: For this DOM: <i>x</i>\n'\n<i>y</i> |
| 107 | // node.textContent = \n'\n and text = ' |
| 108 | // Those newline separators can prevent unnecessary <nowiki/> protection |
| 109 | // if the string begins with one or more newlines before a leading quote. |
| 110 | $origText = $node->textContent; |
| 111 | if ( substr( $origText, 0, 1 ) === "'" ) { |
| 112 | $prev = DiffDOMUtils::previousNonDeletedSibling( $node ); |
| 113 | if ( !$prev ) { |
| 114 | $prev = $node->parentNode; |
| 115 | } |
| 116 | if ( DOMUtils::isQuoteElt( $prev ) ) { |
| 117 | return true; |
| 118 | } |
| 119 | } |
| 120 | |
| 121 | return false; |
| 122 | } |
| 123 | |
| 124 | /** |
| 125 | * @param string $text |
| 126 | * @param array $opts [ 'node' => Node ] |
| 127 | * @return bool |
| 128 | */ |
| 129 | private static function hasTrailingEscapableQuoteChar( string $text, array $opts ): bool { |
| 130 | $node = $opts['node']; |
| 131 | // Use 'node.textContent' to do the tests since it hasn't had newlines |
| 132 | // stripped out from it. |
| 133 | // Ex: For this DOM: <i>x</i>\n'\n<i>y</i> |
| 134 | // node.textContent = \n'\n and text = ' |
| 135 | // Those newline separators can prevent unnecessary <nowiki/> protection |
| 136 | // if the string ends with a trailing quote and then one or more newlines. |
| 137 | $origText = $node->textContent; |
| 138 | if ( substr( $origText, -1 ) === "'" ) { |
| 139 | $next = DiffDOMUtils::nextNonDeletedSibling( $node ); |
| 140 | if ( !$next ) { |
| 141 | $next = $node->parentNode; |
| 142 | } |
| 143 | if ( DOMUtils::isQuoteElt( $next ) ) { |
| 144 | return true; |
| 145 | } |
| 146 | } |
| 147 | |
| 148 | return false; |
| 149 | } |
| 150 | |
| 151 | /** |
| 152 | * SSS FIXME: By doing a DOM walkahead to identify what else is on the current line, |
| 153 | * these heuristics can be improved. Ex: '<i>foo</i> blah blah does not require a |
| 154 | * <nowiki/> after the single quote since we know that there are no other quotes on |
| 155 | * the rest of the line that will be emitted. Similarly, '' does not need a <nowiki> |
| 156 | * wrapper since there are on other quote chars on the line. |
| 157 | * |
| 158 | * This is checking text-node siblings of i/b tags. |
| 159 | * |
| 160 | * @param SerializerState $state |
| 161 | * @param string $text |
| 162 | * @param array $opts [ 'node' => Node ] |
| 163 | * @return string |
| 164 | */ |
| 165 | private static function escapedIBSiblingNodeText( |
| 166 | SerializerState $state, string $text, array $opts |
| 167 | ): string { |
| 168 | // For a sequence of 2+ quote chars, we have to |
| 169 | // fully wrap the sequence in <nowiki>...</nowiki> |
| 170 | // <nowiki/> at the start and end doesn't work. |
| 171 | // |
| 172 | // Ex: ''<i>foo</i> should serialize to <nowiki>''</nowiki>''foo''. |
| 173 | // |
| 174 | // Serializing it to ''<nowiki/>''foo'' breaks html2html semantics |
| 175 | // since it will parse back to <i><meta../></i>foo<i></i> |
| 176 | if ( preg_match( "/''+/", $text ) ) { |
| 177 | // Minimize the length of the string that is wrapped in <nowiki>. |
| 178 | $pieces = explode( "'", $text ); |
| 179 | $first = array_shift( $pieces ); |
| 180 | $last = array_pop( $pieces ); |
| 181 | return $first . "<nowiki>'" . implode( "'", $pieces ) . "'</nowiki>" . $last; |
| 182 | } |
| 183 | |
| 184 | // Check whether the head and/or tail of the text needs <nowiki/> protection. |
| 185 | $out = ''; |
| 186 | if ( self::hasTrailingEscapableQuoteChar( $text, $opts ) ) { |
| 187 | $state->hasQuoteNowikis = true; |
| 188 | $out = $text . '<nowiki/>'; |
| 189 | } |
| 190 | |
| 191 | if ( self::hasLeadingEscapableQuoteChar( $text, $opts ) ) { |
| 192 | $state->hasQuoteNowikis = true; |
| 193 | $out = '<nowiki/>' . ( $out ?: $text ); |
| 194 | } |
| 195 | |
| 196 | return $out; |
| 197 | } |
| 198 | |
| 199 | public function isFirstContentNode( Node $node ): bool { |
| 200 | // Skip deleted-node markers |
| 201 | return DiffDOMUtils::previousNonDeletedSibling( $node ) === null; |
| 202 | } |
| 203 | |
| 204 | /** |
| 205 | * @param Node $liNode |
| 206 | * @param SerializerState $state |
| 207 | * @param string $text |
| 208 | * @param array $opts [ 'node' => Node ] |
| 209 | * @return bool |
| 210 | */ |
| 211 | public function liHandler( |
| 212 | Node $liNode, SerializerState $state, string $text, array $opts |
| 213 | ): bool { |
| 214 | /** @var Node $node */ |
| 215 | $node = $opts['node']; |
| 216 | if ( $node->parentNode !== $liNode ) { |
| 217 | return false; |
| 218 | } |
| 219 | |
| 220 | // For <dt> nodes, ":" trigger nowiki outside of elements |
| 221 | // For first nodes of <li>'s, bullets in sol posn trigger escaping |
| 222 | if ( DOMCompat::nodeName( $liNode ) === 'dt' && str_contains( $text, ':' ) ) { |
| 223 | return true; |
| 224 | } elseif ( preg_match( '/^[#*:;]*$/D', $state->currLine->text ) && |
| 225 | $this->isFirstContentNode( $node ) |
| 226 | ) { |
| 227 | // Wikitext styling might require whitespace insertion after list bullets. |
| 228 | // In those scenarios, presence of bullet-wiktext in the text node is okay. |
| 229 | // Hence the check for /^[#*:;]*$/ above. |
| 230 | return (bool)strspn( $text, '#*:;', 0, 1 ); |
| 231 | } else { |
| 232 | return false; |
| 233 | } |
| 234 | } |
| 235 | |
| 236 | /** |
| 237 | * @param Node $thNode |
| 238 | * @param SerializerState $state |
| 239 | * @param string $text |
| 240 | * @param array $opts [ 'node' => Node ] |
| 241 | * @return bool |
| 242 | */ |
| 243 | public function thHandler( |
| 244 | Node $thNode, SerializerState $state, string $text, array $opts |
| 245 | ): bool { |
| 246 | // {| |
| 247 | // !a<div>!!b</div> |
| 248 | // !c<div>||d</div> |
| 249 | // |} |
| 250 | // |
| 251 | // The <div> will get split across two <th> tags because |
| 252 | // the !! and | has higher precedence in the tokenizer. |
| 253 | // |
| 254 | // So, no matter where in the DOM subtree of the <th> node |
| 255 | // that text shows up in, we have to unconditionally escape |
| 256 | // the !! and | characters. |
| 257 | // |
| 258 | // That is, so long as it serializes to the same line as the |
| 259 | // heading was started. |
| 260 | return preg_match( '/^\s*!/', $state->currLine->text ) && |
| 261 | preg_match( '/^[^\n]*!!|\|/', $text ); |
| 262 | } |
| 263 | |
| 264 | public function mediaOptionHandler( SerializerState $state, string $text ): bool { |
| 265 | return str_contains( $text, '|' ) || preg_match( self::LINKS_ESCAPE_RE, $text ); |
| 266 | } |
| 267 | |
| 268 | public function wikilinkHandler( SerializerState $state, string $text ): bool { |
| 269 | return (bool)preg_match( self::LINKS_ESCAPE_RE, $text ); |
| 270 | } |
| 271 | |
| 272 | public function aHandler( SerializerState $state, string $text ): bool { |
| 273 | return str_contains( $text, ']' ); |
| 274 | } |
| 275 | |
| 276 | /** |
| 277 | * @param Node $tdNode |
| 278 | * @param bool $inWideTD |
| 279 | * @param SerializerState $state |
| 280 | * @param string $text |
| 281 | * @param array $opts [ 'node' => ?Node ] |
| 282 | * @return bool |
| 283 | */ |
| 284 | public function tdHandler( |
| 285 | Node $tdNode, bool $inWideTD, SerializerState $state, string $text, array $opts |
| 286 | ): bool { |
| 287 | $node = $opts['node'] ?? null; |
| 288 | /* |
| 289 | * "|" anywhere in a text node of the <td> subtree can be trouble! |
| 290 | * It is not sufficient to just look at immediate child of <td> |
| 291 | * Try parsing the following table: |
| 292 | * |
| 293 | * {| |
| 294 | * |a''b|c'' |
| 295 | * |} |
| 296 | * |
| 297 | * Similarly, "-" or "+" when emitted after a "|" in sol position |
| 298 | * is trouble, but in addition to showing up as the immediate first |
| 299 | * child of tdNode, they can appear on the leftmost path from |
| 300 | * tdNode as long as the path only has nodes don't emit any wikitext. |
| 301 | * Ex: <td><p>-</p></td>, but not: <td><small>-</small></td> |
| 302 | */ |
| 303 | |
| 304 | // If 'text' is on the same wikitext line as the "|" corresponding |
| 305 | // to the <td> |
| 306 | // * | in a td should be escaped |
| 307 | // * +-} in SOL position (if they show up on the leftmost path with |
| 308 | // only zero-wt-emitting nodes on that path) |
| 309 | if ( !$node || $state->currLine->firstNode === $tdNode ) { |
| 310 | if ( str_contains( $text, '|' ) ) { |
| 311 | return true; |
| 312 | } |
| 313 | if ( !$inWideTD && |
| 314 | $state->currLine->text === '|' && |
| 315 | strspn( $text, '-+}', 0, 1 ) && |
| 316 | $node |
| 317 | ) { |
| 318 | $n = $node; |
| 319 | while ( $n && $n !== $tdNode ) { |
| 320 | if ( !$this->isFirstContentNode( $n ) || |
| 321 | !( $n === $node || WTUtils::isZeroWidthWikitextElt( $n ) ) ) { |
| 322 | return false; |
| 323 | } |
| 324 | $n = $n->parentNode; |
| 325 | } |
| 326 | return true; |
| 327 | } |
| 328 | } |
| 329 | return false; |
| 330 | } |
| 331 | |
| 332 | /** |
| 333 | * Tokenize string and pop EOFTk |
| 334 | * |
| 335 | * @param string $str |
| 336 | * @param bool $sol |
| 337 | * @return array |
| 338 | */ |
| 339 | public function tokenizeStr( string $str, bool $sol ): array { |
| 340 | $tokens = $this->tokenizer->tokenizeSync( $str, [ 'sol' => $sol ] ); |
| 341 | Assert::invariant( |
| 342 | array_pop( $tokens ) instanceof EOFTk, |
| 343 | 'Expected EOF token!' |
| 344 | ); |
| 345 | return $tokens; |
| 346 | } |
| 347 | |
| 348 | public function textCanParseAsLink( Node $node, SerializerState $state, string $text ): bool { |
| 349 | $env = $state->getEnv(); |
| 350 | $env->log( |
| 351 | 'trace/wt-escape', 'link-test-text=', |
| 352 | static function () use ( $text ) { |
| 353 | return PHPUtils::jsonEncode( $text ); |
| 354 | } |
| 355 | ); |
| 356 | |
| 357 | // Strip away extraneous characters after a ]] or a ] |
| 358 | // They are inessential to the test of whether the ]]/] |
| 359 | // will get parsed into a wikilink and only complicate |
| 360 | // the logic (needing to ignore entities, etc.). |
| 361 | $text = preg_replace( '/\][^\]]*$/D', ']', $text, 1 ); |
| 362 | |
| 363 | // text only contains ']' chars. |
| 364 | // Since we stripped everything after ']' above, if a newline is |
| 365 | // present, a link would have to straddle newlines which is not valid. |
| 366 | if ( str_contains( $text, "\n" ) ) { |
| 367 | return false; |
| 368 | } |
| 369 | |
| 370 | $str = $state->currLine->text . $text; |
| 371 | $tokens = $this->tokenizeStr( $str, false ); // sol state is irrelevant here |
| 372 | $n = count( $tokens ); |
| 373 | $lastToken = $tokens[$n - 1]; |
| 374 | |
| 375 | $env->log( 'trace/wt-escape', 'str=', $str, ';tokens=', $tokens ); |
| 376 | |
| 377 | // If 'text' remained outside of any non-string tokens, |
| 378 | // it does not need nowiking. |
| 379 | if ( $lastToken === $text || |
| 380 | ( is_string( $lastToken ) && |
| 381 | $text === substr( $lastToken, -strlen( $text ) ) |
| 382 | ) |
| 383 | ) { |
| 384 | return false; |
| 385 | } |
| 386 | |
| 387 | // Verify that the tokenized links are valid links |
| 388 | $buf = ''; |
| 389 | for ( $i = $n - 1; $i >= 0; $i-- ) { |
| 390 | $t = $tokens[$i]; |
| 391 | if ( is_string( $t ) ) { |
| 392 | $buf = $t . $buf; |
| 393 | } elseif ( $t->getName() === 'wikilink' ) { |
| 394 | $target = $t->getAttributeV( 'href' ); |
| 395 | if ( is_array( $target ) ) { |
| 396 | // FIXME: in theory template expansion *could* make this a link. |
| 397 | return false; |
| 398 | } |
| 399 | if ( $env->isValidLinkTarget( $target ) && |
| 400 | !$env->getSiteConfig()->hasValidProtocol( $target ) |
| 401 | ) { |
| 402 | return true; |
| 403 | } |
| 404 | |
| 405 | // Assumes 'src' will always be present which it seems to be. |
| 406 | // Tests will fail if anything changes in the tokenizer. |
| 407 | $buf = $t->dataParsoid->src . $buf; |
| 408 | } elseif ( $t->getName() === 'extlink' ) { |
| 409 | // Check if the extlink came from a template which in the end |
| 410 | // would not really parse as an extlink. |
| 411 | |
| 412 | $href = $t->getAttributeV( 'href' ); |
| 413 | if ( is_array( $href ) ) { |
| 414 | $href = $href[0]; |
| 415 | } |
| 416 | |
| 417 | if ( !TokenUtils::isTemplateToken( $href ) ) { |
| 418 | // Not a template and a real href => needs nowiking |
| 419 | if ( is_string( $href ) && preg_match( '#https?://#', $href ) ) { |
| 420 | return true; |
| 421 | } |
| 422 | } else { |
| 423 | while ( $node ) { |
| 424 | $node = DiffDOMUtils::previousNonSepSibling( $node ); |
| 425 | if ( $node && WTUtils::isFirstEncapsulationWrapperNode( $node ) ) { |
| 426 | // FIXME: This is not entirely correct. |
| 427 | // Assumes that extlink content doesn't have templates. |
| 428 | // Solution: Count # of non-nested templates encountered |
| 429 | // and skip over intermediate templates. |
| 430 | // var content = t.getAttribute('mw:content'); |
| 431 | // var n = intermediateNonNestedTemplates(content); |
| 432 | break; |
| 433 | } |
| 434 | } |
| 435 | |
| 436 | if ( $node instanceof Element && DOMCompat::nodeName( $node ) === 'a' && |
| 437 | $node->textContent === DOMCompat::getAttribute( $node, 'href' ) |
| 438 | ) { |
| 439 | // The template expands to an url link => needs nowiking |
| 440 | return true; |
| 441 | } |
| 442 | } |
| 443 | |
| 444 | // Since this will not parse to a real extlink, |
| 445 | // update buf with the wikitext src for this token. |
| 446 | $tsr = $t->dataParsoid->tsr; |
| 447 | $buf = $tsr->substr( $str ) . $buf; |
| 448 | } else { |
| 449 | // We have no other smarts => be conservative. |
| 450 | return true; |
| 451 | } |
| 452 | |
| 453 | if ( $text === substr( $buf, -strlen( $text ) ) ) { |
| 454 | // 'text' emerged unscathed |
| 455 | return false; |
| 456 | } |
| 457 | } |
| 458 | |
| 459 | // We couldn't prove safety of skipping nowiki-ing. |
| 460 | return true; |
| 461 | } |
| 462 | |
| 463 | private function hasWikitextTokens( |
| 464 | SerializerState $state, bool $onNewline, string $text |
| 465 | ): bool { |
| 466 | $env = $state->getEnv(); |
| 467 | $env->log( |
| 468 | 'trace/wt-escape', 'nl:', $onNewline, ':text=', |
| 469 | static function () use ( $text ) { |
| 470 | return PHPUtils::jsonEncode( $text ); |
| 471 | } |
| 472 | ); |
| 473 | |
| 474 | // tokenize the text |
| 475 | $sol = $onNewline && !( $state->inIndentPre || $state->inPHPBlock ); |
| 476 | |
| 477 | // If we're inside a <pre>, we need to add an extra space after every |
| 478 | // newline so that the tokenizer correctly parses all tokens in a pre |
| 479 | // instead of just the first one. See T95794. |
| 480 | if ( $state->inIndentPre ) { |
| 481 | $text = str_replace( "\n", "\n ", $text ); |
| 482 | } |
| 483 | |
| 484 | $tokens = $this->tokenizeStr( $text, $sol ); |
| 485 | |
| 486 | // If the token stream has a TagTk, SelfclosingTagTk, EndTagTk or CommentTk |
| 487 | // then this text needs escaping! |
| 488 | $numEntities = 0; |
| 489 | foreach ( $tokens as $t ) { |
| 490 | $env->log( |
| 491 | 'trace/wt-escape', 'T:', |
| 492 | static function () use ( $t ) { |
| 493 | return PHPUtils::jsonEncode( $t ); |
| 494 | } |
| 495 | ); |
| 496 | |
| 497 | $tc = TokenUtils::getTokenType( $t ); |
| 498 | |
| 499 | // Ignore html tags that aren't allowed as literals in wikitext |
| 500 | if ( TokenUtils::isHTMLTag( $t ) ) { |
| 501 | if ( |
| 502 | TokenUtils::matchTypeOf( $t, '#^mw:Extension(/|$)#' ) && |
| 503 | ( $this->extName !== $t->getAttributeV( 'name' ) ) |
| 504 | ) { |
| 505 | return true; |
| 506 | } |
| 507 | |
| 508 | // Always escape isolated extension tags (T59469). Consider this: |
| 509 | // echo "<ref>foo<p></ref></p>" | node parse --html2wt |
| 510 | // The <ref> and </ref> tag-like text is spread across the DOM, and in |
| 511 | // the worst case can be anywhere. So, we conservatively escape these |
| 512 | // elements always (which can lead to excessive nowiki-escapes in some |
| 513 | // cases, but is always safe). |
| 514 | if ( ( $tc === 'TagTk' || $tc === 'EndTagTk' ) && |
| 515 | $env->getSiteConfig()->isExtensionTag( mb_strtolower( $t->getName() ) ) |
| 516 | ) { |
| 517 | return true; |
| 518 | } |
| 519 | |
| 520 | // If the tag is one that's allowed in wikitext, we need to escape |
| 521 | // it inside <nowiki>s, because a text node nodeValue always returns |
| 522 | // non-escaped entities (that is, converts "<h2>" to "<h2>"). |
| 523 | // TODO: We should also do this for <a> tags because even if they |
| 524 | // aren't allowed in wikitext and thus don't need to be escaped, the |
| 525 | // result can be confusing for editors. However, doing it here in a |
| 526 | // simple way interacts badly with normal link escaping, so it's |
| 527 | // left for later. |
| 528 | if ( isset( Consts::$Sanitizer['AllowedLiteralTags'][mb_strtolower( $t->getName() )] ) ) { |
| 529 | return true; |
| 530 | } else { |
| 531 | continue; |
| 532 | } |
| 533 | } |
| 534 | |
| 535 | if ( $tc === 'SelfclosingTagTk' ) { |
| 536 | // * Ignore RFC/ISBN/PMID tokens when those are encountered in the |
| 537 | // context of another link's content -- those are not parsed to |
| 538 | // ext-links in that context. (T109371) |
| 539 | if ( ( $t->getName() === 'extlink' || $t->getName() === 'wikilink' ) && |
| 540 | ( $t->dataParsoid->stx ?? null ) === 'magiclink' && |
| 541 | ( $state->inAttribute || $state->inLink ) ) { |
| 542 | continue; |
| 543 | } |
| 544 | |
| 545 | // Ignore url links in attributes (href, mostly) |
| 546 | // since they are not in danger of being autolink-ified there. |
| 547 | if ( $t->getName() === 'urllink' && ( $state->inAttribute || $state->inLink ) ) { |
| 548 | continue; |
| 549 | } |
| 550 | |
| 551 | if ( $t->getName() === 'wikilink' ) { |
| 552 | if ( $env->isValidLinkTarget( $t->getAttributeV( 'href' ) ?? '' ) ) { |
| 553 | return true; |
| 554 | } else { |
| 555 | continue; |
| 556 | } |
| 557 | } |
| 558 | |
| 559 | return true; |
| 560 | } |
| 561 | |
| 562 | if ( $state->inCaption && $tc === 'TagTk' && $t->getName() === 'listItem' ) { |
| 563 | continue; |
| 564 | } |
| 565 | |
| 566 | if ( $tc === 'TagTk' ) { |
| 567 | // Ignore mw:Entity tokens |
| 568 | if ( $t->getName() === 'span' && TokenUtils::hasTypeOf( $t, 'mw:Entity' ) ) { |
| 569 | $numEntities++; |
| 570 | continue; |
| 571 | } |
| 572 | |
| 573 | // Ignore table tokens outside of tables |
| 574 | if ( in_array( $t->getName(), [ 'caption', 'td', 'tr', 'th' ], true ) && |
| 575 | !TokenUtils::isHTMLTag( $t ) && |
| 576 | $state->wikiTableNesting === 0 |
| 577 | ) { |
| 578 | continue; |
| 579 | } |
| 580 | |
| 581 | // Headings have both SOL and EOL requirements. This tokenization |
| 582 | // here only verifies SOL requirements, not EOL requirements. |
| 583 | // So, record this information so that we can strip unnecessary |
| 584 | // nowikis after the fact. |
| 585 | if ( preg_match( '/^h\d$/D', $t->getName() ) ) { |
| 586 | $state->hasHeadingEscapes = true; |
| 587 | } |
| 588 | |
| 589 | return true; |
| 590 | } |
| 591 | |
| 592 | if ( $tc === 'EndTagTk' ) { |
| 593 | // Ignore mw:Entity tokens |
| 594 | if ( $numEntities > 0 && $t->getName() === 'span' ) { |
| 595 | $numEntities--; |
| 596 | continue; |
| 597 | } |
| 598 | // Ignore heading tokens |
| 599 | if ( preg_match( '/^h\d$/D', $t->getName() ) ) { |
| 600 | continue; |
| 601 | } |
| 602 | |
| 603 | // Ignore table tokens outside of tables |
| 604 | if ( isset( ( [ 'caption' => 1, 'table' => 1 ] )[$t->getName()] ) && |
| 605 | $state->wikiTableNesting === 0 |
| 606 | ) { |
| 607 | continue; |
| 608 | } |
| 609 | |
| 610 | // </br>! |
| 611 | if ( mb_strtolower( $t->getName() ) === 'br' ) { |
| 612 | continue; |
| 613 | } |
| 614 | |
| 615 | return true; |
| 616 | } |
| 617 | } |
| 618 | |
| 619 | return false; |
| 620 | } |
| 621 | |
| 622 | private static function nowikiWrap( |
| 623 | string $str, bool $close, bool &$inNowiki, bool &$nowikisAdded, string &$buf |
| 624 | ): void { |
| 625 | if ( !$inNowiki ) { |
| 626 | $buf .= '<nowiki>'; |
| 627 | $inNowiki = true; |
| 628 | $nowikisAdded = true; |
| 629 | } |
| 630 | $buf .= $str; |
| 631 | if ( $close ) { |
| 632 | $buf .= '</nowiki>'; |
| 633 | $inNowiki = false; |
| 634 | } |
| 635 | } |
| 636 | |
| 637 | /** |
| 638 | * This function attempts to wrap smallest escapable units into |
| 639 | * nowikis (which can potentially add multiple nowiki pairs in a |
| 640 | * single string). The idea here is that since this should all be |
| 641 | * text, anything that tokenizes to another construct needs to be |
| 642 | * wrapped. |
| 643 | * |
| 644 | * Full-wrapping is enabled if the string is being escaped within |
| 645 | * context-specific handlers where the tokenization context might |
| 646 | * be different from what we use in this code. |
| 647 | * |
| 648 | * @param SerializerState $state |
| 649 | * @param bool $sol |
| 650 | * @param string $origText |
| 651 | * @param bool $fullWrap |
| 652 | * @param bool $dontWrapIfUnnecessary |
| 653 | * @return string |
| 654 | */ |
| 655 | public function escapedText( |
| 656 | SerializerState $state, bool $sol, string $origText, |
| 657 | bool $fullWrap = false, bool $dontWrapIfUnnecessary = false |
| 658 | ): string { |
| 659 | Assert::invariant( |
| 660 | preg_match( '/^(.*?)((?:\r?\n)*)$/sD', $origText, $match ), |
| 661 | "Escaped text matching failed: {$origText}" |
| 662 | ); |
| 663 | |
| 664 | $text = $match[1]; |
| 665 | $nls = $match[2]; |
| 666 | |
| 667 | if ( $fullWrap ) { |
| 668 | return '<nowiki>' . $text . '</nowiki>' . $nls; |
| 669 | } |
| 670 | |
| 671 | $buf = ''; |
| 672 | $inNowiki = false; |
| 673 | $nowikisAdded = false; |
| 674 | // These token types don't come with a closing tag |
| 675 | $tokensWithoutClosingTag = PHPUtils::makeSet( [ 'listItem', 'td', 'tr' ] ); |
| 676 | |
| 677 | // reverse escaping nowiki tags |
| 678 | // we do this so that they tokenize as nowikis |
| 679 | // instead of entity enclosed text |
| 680 | $text = preg_replace( '#<(/?nowiki\s*/?\s*)>#i', '<$1>', $text ); |
| 681 | |
| 682 | $tokens = $this->tokenizeStr( $text, $sol ); |
| 683 | |
| 684 | foreach ( $tokens as $t ) { |
| 685 | if ( is_string( $t ) ) { |
| 686 | if ( strlen( $t ) > 0 ) { |
| 687 | $t = WTSUtils::escapeNowikiTags( $t ); |
| 688 | if ( !$inNowiki && ( ( $sol && $t[0] === ' ' ) || str_contains( $t, "\n " ) ) ) { |
| 689 | $x = preg_split( '/(^|\n) /', $t, -1, PREG_SPLIT_DELIM_CAPTURE ); |
| 690 | $buf .= $x[0]; |
| 691 | $lastIndexX = count( $x ) - 1; |
| 692 | for ( $k = 1; $k < $lastIndexX; $k += 2 ) { |
| 693 | $buf .= $x[$k]; |
| 694 | if ( $k !== 1 || $x[$k] === "\n" || $sol ) { |
| 695 | self::nowikiWrap( ' ', true, $inNowiki, $nowikisAdded, $buf ); |
| 696 | } else { |
| 697 | $buf .= ' '; |
| 698 | } |
| 699 | $buf .= $x[$k + 1]; |
| 700 | } |
| 701 | } else { |
| 702 | $buf .= $t; |
| 703 | } |
| 704 | $sol = false; |
| 705 | } |
| 706 | continue; |
| 707 | } |
| 708 | |
| 709 | $tsr = $t->dataParsoid->tsr ?? null; |
| 710 | if ( !( $tsr instanceof SourceRange ) ) { |
| 711 | $env = $state->getEnv(); |
| 712 | $env->log( |
| 713 | 'error/html2wt/escapeNowiki', |
| 714 | 'Missing tsr for token ', |
| 715 | PHPUtils::jsonEncode( $t ), |
| 716 | 'while processing text ', |
| 717 | $text |
| 718 | ); |
| 719 | |
| 720 | // Bail and wrap the whole thing in a nowiki |
| 721 | // if we have missing information. |
| 722 | // Use match[1] since text has been clobbered above. |
| 723 | return '<nowiki>' . $match[1] . '</nowiki>' . $nls; |
| 724 | } |
| 725 | |
| 726 | // Now put back the escaping we removed above |
| 727 | $tSrc = WTSUtils::escapeNowikiTags( $tsr->substr( $text ) ); |
| 728 | switch ( TokenUtils::getTokenType( $t ) ) { |
| 729 | case 'NlTk': |
| 730 | $buf .= $tSrc; |
| 731 | $sol = true; |
| 732 | break; |
| 733 | case 'CommentTk': |
| 734 | // Comments are sol-transparent |
| 735 | $buf .= $tSrc; |
| 736 | break; |
| 737 | case 'TagTk': |
| 738 | // Treat tokens with missing tags as self-closing tokens |
| 739 | // for the purpose of minimal nowiki escaping |
| 740 | self::nowikiWrap( |
| 741 | $tSrc, |
| 742 | isset( $tokensWithoutClosingTag[$t->getName()] ), |
| 743 | $inNowiki, |
| 744 | $nowikisAdded, |
| 745 | $buf |
| 746 | ); |
| 747 | $sol = false; |
| 748 | break; |
| 749 | case 'EndTagTk': |
| 750 | self::nowikiWrap( $tSrc, true, $inNowiki, $nowikisAdded, $buf ); |
| 751 | $sol = false; |
| 752 | break; |
| 753 | case 'SelfclosingTagTk': |
| 754 | if ( $t->getName() !== 'meta' || |
| 755 | !TokenUtils::hasTypeOf( $t, 'mw:EmptyLine' ) |
| 756 | ) { |
| 757 | // Don't bother with marker or empty-line metas |
| 758 | self::nowikiWrap( $tSrc, true, $inNowiki, $nowikisAdded, $buf ); |
| 759 | } |
| 760 | $sol = false; |
| 761 | break; |
| 762 | } |
| 763 | } |
| 764 | |
| 765 | // close any unclosed nowikis |
| 766 | if ( $inNowiki ) { |
| 767 | $buf .= '</nowiki>'; |
| 768 | } |
| 769 | |
| 770 | // Make sure nowiki is always added |
| 771 | // Ex: "foo]]" won't tokenize into tags at all |
| 772 | if ( !$nowikisAdded && !$dontWrapIfUnnecessary ) { |
| 773 | $buf = ''; |
| 774 | self::nowikiWrap( $text, true, $inNowiki, $nowikisAdded, $buf ); |
| 775 | } |
| 776 | |
| 777 | $buf .= $nls; |
| 778 | return $buf; |
| 779 | } |
| 780 | |
| 781 | /** |
| 782 | * @param SerializerState $state |
| 783 | * @param string $text |
| 784 | * @param array $opts [ 'node' => Node, 'inMultilineMode' => ?bool, 'isLastChild' => ?bool ] |
| 785 | * @return string |
| 786 | */ |
| 787 | public function escapeWikitext( SerializerState $state, string $text, array $opts ): string { |
| 788 | $env = $state->getEnv(); |
| 789 | $env->log( |
| 790 | 'trace/wt-escape', 'EWT:', |
| 791 | static function () use ( $text ) { |
| 792 | return PHPUtils::jsonEncode( $text ); |
| 793 | } |
| 794 | ); |
| 795 | |
| 796 | /* ----------------------------------------------------------------- |
| 797 | * General strategy: If a substring requires escaping, we can escape |
| 798 | * the entire string without further analysis of the rest of the string. |
| 799 | * ----------------------------------------------------------------- */ |
| 800 | |
| 801 | $hasMagicWord = preg_match( '/(^|\W)(RFC|ISBN|PMID)\s/', $text ); |
| 802 | $hasAutolink = $env->getSiteConfig()->findValidProtocol( $text ); |
| 803 | $fullCheckNeeded = !$state->inLink && ( $hasMagicWord || $hasAutolink ); |
| 804 | $hasQuoteChar = false; |
| 805 | $indentPreUnsafe = false; |
| 806 | $hasNonQuoteEscapableChars = false; |
| 807 | $indentPreSafeMode = $state->inIndentPre || $state->inPHPBlock; |
| 808 | $sol = $state->onSOL && !$indentPreSafeMode; |
| 809 | |
| 810 | // Fast path for special protected characters. |
| 811 | if ( $state->protect && preg_match( $state->protect, $text ) ) { |
| 812 | return $this->escapedText( $state, $sol, $text ); |
| 813 | } |
| 814 | |
| 815 | if ( !$fullCheckNeeded ) { |
| 816 | $hasQuoteChar = str_contains( $text, "'" ); |
| 817 | $indentPreUnsafe = !$indentPreSafeMode && ( |
| 818 | preg_match( '/\n +[^\r\n]*?\S+/', $text ) || |
| 819 | $sol && preg_match( '/^ +[^\r\n]*?\S+/', $text ) |
| 820 | ); |
| 821 | $hasNonQuoteEscapableChars = preg_match( '/[<>\[\]\-\+\|!=#\*:;~{}]|__[^_]*__/', $text ); |
| 822 | $hasLanguageConverter = preg_match( '/-\{|\}-/', $text ); |
| 823 | if ( $hasLanguageConverter ) { |
| 824 | $fullCheckNeeded = true; |
| 825 | } |
| 826 | } |
| 827 | |
| 828 | // Quick check for the common case (useful to kill a majority of requests) |
| 829 | // |
| 830 | // Pure white-space or text without wt-special chars need not be analyzed |
| 831 | if ( !$fullCheckNeeded && !$hasQuoteChar && !$indentPreUnsafe && !$hasNonQuoteEscapableChars ) { |
| 832 | $env->log( 'trace/wt-escape', '---No-checks needed---' ); |
| 833 | return $text; |
| 834 | } |
| 835 | |
| 836 | // Context-specific escape handler |
| 837 | $wteHandler = PHPUtils::lastItem( $state->wteHandlerStack ); |
| 838 | if ( $wteHandler && $wteHandler( $state, $text, $opts ) ) { |
| 839 | $env->log( 'trace/wt-escape', '---Context-specific escape handler---' ); |
| 840 | return $this->escapedText( $state, false, $text, true ); |
| 841 | } |
| 842 | |
| 843 | // Quote-escape test |
| 844 | if ( str_contains( $text, "''" ) || |
| 845 | self::hasLeadingEscapableQuoteChar( $text, $opts ) || |
| 846 | self::hasTrailingEscapableQuoteChar( $text, $opts ) |
| 847 | ) { |
| 848 | // Check if we need full-wrapping <nowiki>..</nowiki> |
| 849 | // or selective <nowiki/> escaping for quotes. |
| 850 | if ( $fullCheckNeeded || |
| 851 | $indentPreUnsafe || |
| 852 | ( $hasNonQuoteEscapableChars && |
| 853 | $this->hasWikitextTokens( $state, $sol, $text ) |
| 854 | ) |
| 855 | ) { |
| 856 | $env->log( 'trace/wt-escape', '---quotes: escaping text---' ); |
| 857 | // If the reason for full wrap is that the text contains non-quote |
| 858 | // escapable chars, it's still possible to minimize the contents |
| 859 | // of the <nowiki> (T71950). |
| 860 | return $this->escapedText( $state, $sol, $text ); |
| 861 | } else { |
| 862 | $quoteEscapedText = self::escapedIBSiblingNodeText( $state, $text, $opts ); |
| 863 | if ( $quoteEscapedText ) { |
| 864 | $env->log( 'trace/wt-escape', '---sibling of i/b tag---' ); |
| 865 | return $quoteEscapedText; |
| 866 | } |
| 867 | } |
| 868 | } |
| 869 | |
| 870 | // Template and template-arg markers are escaped unconditionally! |
| 871 | // Conditional escaping requires matching brace pairs and knowledge |
| 872 | // of whether we are in template arg context or not. |
| 873 | if ( preg_match( '/\{\{\{|\{\{|\}\}\}|\}\}/', $text ) ) { |
| 874 | $env->log( 'trace/wt-escape', '---Unconditional: transclusion chars---' ); |
| 875 | return $this->escapedText( $state, false, $text ); |
| 876 | } |
| 877 | |
| 878 | // Once we eliminate the possibility of multi-line tokens, split the text |
| 879 | // around newlines and escape each line separately. |
| 880 | if ( preg_match( '/\n./', $text ) ) { |
| 881 | $env->log( 'trace/wt-escape', '-- <multi-line-escaping-mode> --' ); |
| 882 | // We've already processed the full string in a context-specific handler. |
| 883 | // No more additional processing required. So, push/pop a null handler. |
| 884 | $state->wteHandlerStack[] = null; |
| 885 | |
| 886 | $tmp = []; |
| 887 | foreach ( explode( "\n", $text ) as $i => $line ) { |
| 888 | if ( $i > 0 ) { |
| 889 | // Update state |
| 890 | $state->onSOL = true; |
| 891 | $state->currLine->text = ''; |
| 892 | $opts['inMultilineMode'] = true; |
| 893 | } |
| 894 | $tmp[] = $this->escapeWikitext( $state, $line, $opts ); |
| 895 | } |
| 896 | $ret = implode( "\n", $tmp ); |
| 897 | |
| 898 | array_pop( $state->wteHandlerStack ); |
| 899 | |
| 900 | // If nothing changed, check if the original multiline string has |
| 901 | // any wikitext tokens (ex: multi-line html tags <div\n>foo</div\n>). |
| 902 | if ( $ret === $text && $this->hasWikitextTokens( $state, $sol, $text ) ) { |
| 903 | $env->log( 'trace/wt-escape', '---Found multi-line wt tokens---' ); |
| 904 | $ret = $this->escapedText( $state, $sol, $text ); |
| 905 | } |
| 906 | |
| 907 | $env->log( 'trace/wt-escape', '-- </multi-line-escaping-mode> --' ); |
| 908 | return $ret; |
| 909 | } |
| 910 | |
| 911 | $env->log( |
| 912 | 'trace/wt-escape', 'SOL:', $sol, |
| 913 | static function () use ( $text ) { |
| 914 | return PHPUtils::jsonEncode( $text ); |
| 915 | } |
| 916 | ); |
| 917 | |
| 918 | $hasTildes = preg_match( '/~{3,5}/', $text ); |
| 919 | if ( !$fullCheckNeeded && !$hasTildes ) { |
| 920 | // {{, {{{, }}}, }} are handled above. |
| 921 | // Test 1: '', [], <>, __FOO__ need escaping wherever they occur |
| 922 | // = needs escaping in end-of-line context |
| 923 | // Test 2: {|, |}, ||, |-, |+, , *#:;, ----, =*= need escaping only in SOL context. |
| 924 | if ( !$sol && !preg_match( "/''|[<>]|\\[.*\\]|\\]|(=[ ]*(\\n|$))|__[^_]*__/", $text ) ) { |
| 925 | // It is not necessary to test for an unmatched opening bracket ([) |
| 926 | // as long as we always escape an unmatched closing bracket (]). |
| 927 | $env->log( 'trace/wt-escape', '---Not-SOL and safe---' ); |
| 928 | return $text; |
| 929 | } |
| 930 | |
| 931 | // Quick checks when on a newline |
| 932 | // + can only occur as "|+" and - can only occur as "|-" or ---- |
| 933 | if ( $sol && !preg_match( '/(^|\n)[ #*:;=]|[<\[\]>\|\'!]|\-\-\-\-|__[^_]*__/', $text ) ) { |
| 934 | $env->log( 'trace/wt-escape', '---SOL and safe---' ); |
| 935 | return $text; |
| 936 | } |
| 937 | } |
| 938 | |
| 939 | // The front-end parser eliminated pre-tokens in the tokenizer |
| 940 | // and moved them to a stream handler. So, we always conservatively |
| 941 | // escape text with ' ' in sol posn with one caveat: |
| 942 | // * and when the current line has block tokens |
| 943 | if ( $indentPreUnsafe && |
| 944 | ( !self::hasBlocksOnLine( $state->currLine->firstNode, true ) || |
| 945 | !empty( $opts['inMultilineMode'] ) |
| 946 | ) |
| 947 | ) { |
| 948 | $env->log( 'trace/wt-escape', '---SOL and pre---' ); |
| 949 | $state->hasIndentPreNowikis = true; |
| 950 | return $this->escapedText( $state, $sol, $text ); |
| 951 | } |
| 952 | |
| 953 | // escape nowiki tags |
| 954 | $text = WTSUtils::escapeNowikiTags( $text ); |
| 955 | |
| 956 | // Use the tokenizer to see if we have any wikitext tokens |
| 957 | // |
| 958 | // Ignores entities |
| 959 | if ( $hasTildes ) { |
| 960 | $env->log( 'trace/wt-escape', '---Found tildes---' ); |
| 961 | return $this->escapedText( $state, $sol, $text ); |
| 962 | } elseif ( $this->hasWikitextTokens( $state, $sol, $text ) ) { |
| 963 | $env->log( 'trace/wt-escape', '---Found WT tokens---' ); |
| 964 | return $this->escapedText( $state, $sol, $text ); |
| 965 | } elseif ( preg_match( '/[^\[]*\]/', $text ) && |
| 966 | $this->textCanParseAsLink( $opts['node'], $state, $text ) |
| 967 | ) { |
| 968 | // we have an closing bracket, and |
| 969 | // - the text will get parsed as a link in |
| 970 | $env->log( 'trace/wt-escape', '---Links: complex single-line test---' ); |
| 971 | return $this->escapedText( $state, $sol, $text ); |
| 972 | } elseif ( !empty( $opts['isLastChild'] ) && substr( $text, -1 ) === '=' ) { |
| 973 | // 1. we have an open heading char, and |
| 974 | // - text ends in a '=' |
| 975 | // - text comes from the last child |
| 976 | preg_match( '/^h(\d)/', DOMCompat::nodeName( $state->currLine->firstNode ), $headingMatch ); |
| 977 | if ( $headingMatch ) { |
| 978 | $n = intval( $headingMatch[1] ); |
| 979 | if ( ( $state->currLine->text . $text )[$n] === '=' ) { |
| 980 | // The first character after the heading wikitext is/will be a '='. |
| 981 | // So, the trailing '=' can change semantics if it is not nowikied. |
| 982 | $env->log( 'trace/wt-escape', '---Heading: complex single-line test---' ); |
| 983 | return $this->escapedText( $state, $sol, $text ); |
| 984 | } else { |
| 985 | return $text; |
| 986 | } |
| 987 | } elseif ( strlen( $state->currLine->text ) > 0 && $state->currLine->text[0] === '=' ) { |
| 988 | $env->log( 'trace/wt-escape', '---Text-as-heading: complex single-line test---' ); |
| 989 | return $this->escapedText( $state, $sol, $text ); |
| 990 | } else { |
| 991 | return $text; |
| 992 | } |
| 993 | } else { |
| 994 | $env->log( 'trace/wt-escape', '---All good!---' ); |
| 995 | return $text; |
| 996 | } |
| 997 | } |
| 998 | |
| 999 | /** |
| 1000 | * @param string $str |
| 1001 | * @param bool $isLast |
| 1002 | * @param bool $checkNowiki |
| 1003 | * @param string &$buf |
| 1004 | * @param bool &$openNowiki |
| 1005 | * @param bool $isTemplate |
| 1006 | * @param bool &$serializeAsNamed |
| 1007 | * @param array $opts [ 'numPositionalArgs' => int, 'argPositionalIndex' => int, 'type' => string, |
| 1008 | * 'numArgs' => int, 'argIndex' => int ] |
| 1009 | */ |
| 1010 | private static function appendStr( |
| 1011 | string $str, bool $isLast, bool $checkNowiki, string &$buf, bool &$openNowiki, |
| 1012 | bool $isTemplate, bool &$serializeAsNamed, array $opts |
| 1013 | ): void { |
| 1014 | if ( !$checkNowiki ) { |
| 1015 | if ( $openNowiki ) { |
| 1016 | $buf .= '</nowiki>'; |
| 1017 | $openNowiki = false; |
| 1018 | } |
| 1019 | $buf .= $str; |
| 1020 | return; |
| 1021 | } |
| 1022 | |
| 1023 | // '=' is not allowed in positional parameters. We can either |
| 1024 | // nowiki escape it or convert the named parameter into a |
| 1025 | // positional param to avoid the escaping. |
| 1026 | if ( $isTemplate && !$serializeAsNamed && str_contains( $str, '=' ) ) { |
| 1027 | // In certain situations, it is better to add a nowiki escape |
| 1028 | // rather than convert this to a named param. |
| 1029 | // |
| 1030 | // Ex: Consider: {{funky-tpl|a|b|c|d|e|f|g|h}} |
| 1031 | // |
| 1032 | // If an editor changes 'a' to 'f=oo' and we convert it to |
| 1033 | // a named param 1=f=oo, we are effectively converting all |
| 1034 | // the later params into named params as well and we get |
| 1035 | // {{funky-tpl|1=f=oo|2=b|3=c|...|8=h}} instead of |
| 1036 | // {{funky-tpl|<nowiki>f=oo</nowiki>|b|c|...|h}} |
| 1037 | // |
| 1038 | // The latter is better in this case. This is a real problem |
| 1039 | // in production. |
| 1040 | // |
| 1041 | // For now, we use a simple heuristic below and can be |
| 1042 | // refined later, if necessary |
| 1043 | // |
| 1044 | // 1. Either there were no original positional args |
| 1045 | // 2. Or, only the last positional arg uses '=' |
| 1046 | if ( $opts['numPositionalArgs'] === 0 || |
| 1047 | $opts['numPositionalArgs'] === $opts['argPositionalIndex'] |
| 1048 | ) { |
| 1049 | $serializeAsNamed = true; |
| 1050 | } |
| 1051 | } |
| 1052 | |
| 1053 | // Count how many reasons for nowiki |
| 1054 | $needNowikiCount = 0; |
| 1055 | $neededSubstitution = null; |
| 1056 | // Protect against unmatched pairs of braces and brackets, as they |
| 1057 | // should never appear in template arguments. |
| 1058 | $bracketPairStrippedStr = preg_replace( |
| 1059 | '/\[\[([^\[\]]*)\]\]|\{\{([^\{\}]*)\}\}|-\{([^\{\}]*)\}-/', |
| 1060 | '_$1_', |
| 1061 | $str |
| 1062 | ); |
| 1063 | if ( preg_match( '/\{\{|\}\}|\[\[|\]\]|-\{/', $bracketPairStrippedStr ) ) { |
| 1064 | $needNowikiCount++; |
| 1065 | } |
| 1066 | if ( $opts['type'] !== 'templatearg' && !$serializeAsNamed && str_contains( $str, '=' ) ) { |
| 1067 | $needNowikiCount++; |
| 1068 | } |
| 1069 | if ( $opts['argIndex'] === $opts['numArgs'] && $isLast && str_ends_with( $str, '}' ) ) { |
| 1070 | // If this is the last part of the last argument, we need to protect |
| 1071 | // against an ending }, as it would get confused with the template ending }}. |
| 1072 | $needNowikiCount++; |
| 1073 | $neededSubstitution = [ '/(\})$/D', '<nowiki>}</nowiki>' ]; |
| 1074 | } |
| 1075 | if ( str_contains( $str, '|' ) ) { |
| 1076 | // If there's an unprotected |, guard it so it doesn't get confused |
| 1077 | // with the beginning of a different parameter. |
| 1078 | $needNowikiCount++; |
| 1079 | $neededSubstitution = [ '/\|/', '{{!}}' ]; |
| 1080 | } |
| 1081 | |
| 1082 | // Now, if arent' already in a <nowiki> and there's only one reason to |
| 1083 | // protect, avoid guarding too much text by just substituting. |
| 1084 | if ( !$openNowiki && $needNowikiCount === 1 && $neededSubstitution ) { |
| 1085 | $str = preg_replace( $neededSubstitution[0], $neededSubstitution[1], $str ); |
| 1086 | $needNowikiCount = false; |
| 1087 | } |
| 1088 | if ( !$openNowiki && $needNowikiCount ) { |
| 1089 | $buf .= '<nowiki>'; |
| 1090 | $openNowiki = true; |
| 1091 | } |
| 1092 | if ( !$needNowikiCount && $openNowiki ) { |
| 1093 | $buf .= '</nowiki>'; |
| 1094 | $openNowiki = false; |
| 1095 | } |
| 1096 | $buf .= $str; |
| 1097 | } |
| 1098 | |
| 1099 | /** |
| 1100 | * General strategy: |
| 1101 | * |
| 1102 | * Tokenize the arg wikitext. Anything that parses as tags |
| 1103 | * are good and we need not bother with those. Check for harmful |
| 1104 | * characters `[[]]{{}}` or additionally `=` in positional parameters and escape |
| 1105 | * those fragments since these characters could change semantics of the entire |
| 1106 | * template transclusion. |
| 1107 | * |
| 1108 | * This function makes a couple of assumptions: |
| 1109 | * |
| 1110 | * 1. The tokenizer sets tsr on all non-string tokens. |
| 1111 | * 2. The tsr on TagTk and EndTagTk corresponds to the |
| 1112 | * width of the opening and closing wikitext tags and not |
| 1113 | * the entire DOM range they span in the end. |
| 1114 | * |
| 1115 | * @param string $arg |
| 1116 | * @param array $opts [ 'serializeAsNamed' => bool, 'numPositionalArgs' => int, |
| 1117 | * 'argPositionalIndex' => int, 'type' => string, 'numArgs' => int, 'argIndex' => int ] |
| 1118 | * @return array |
| 1119 | */ |
| 1120 | public function escapeTplArgWT( string $arg, array $opts ): array { |
| 1121 | $env = $this->env; |
| 1122 | $serializeAsNamed = $opts['serializeAsNamed']; |
| 1123 | $buf = ''; |
| 1124 | $openNowiki = false; |
| 1125 | $isTemplate = $opts['type'] === 'template'; |
| 1126 | |
| 1127 | $tokens = $this->tokenizeStr( $arg, false ); |
| 1128 | |
| 1129 | for ( $i = 0, $n = count( $tokens ); $i < $n; $i++ ) { |
| 1130 | $t = $tokens[$i]; |
| 1131 | $last = $i === $n - 1; |
| 1132 | |
| 1133 | // For mw:Entity spans, the opening and closing tags have 0 width |
| 1134 | // and the enclosed content is the decoded entity. Hence the |
| 1135 | // special case to serialize back the entity's source. |
| 1136 | if ( $t instanceof TagTk ) { |
| 1137 | $da = $t->dataParsoid; |
| 1138 | if ( TokenUtils::matchTypeOf( $t, '#^mw:(Placeholder|Entity)(/|$)#' ) ) { |
| 1139 | $i += 2; |
| 1140 | $width = $tokens[$i]->dataParsoid->tsr->end - $da->tsr->start; |
| 1141 | self::appendStr( |
| 1142 | substr( $arg, $da->tsr->start, $width ), |
| 1143 | $last, |
| 1144 | false, |
| 1145 | $buf, |
| 1146 | $openNowiki, |
| 1147 | $isTemplate, |
| 1148 | $serializeAsNamed, |
| 1149 | $opts |
| 1150 | ); |
| 1151 | continue; |
| 1152 | } elseif ( TokenUtils::hasTypeOf( $t, 'mw:Nowiki' ) ) { |
| 1153 | $i++; |
| 1154 | while ( $i < $n && |
| 1155 | ( !$tokens[$i] instanceof EndTagTk || |
| 1156 | !TokenUtils::hasTypeOf( $tokens[$i], 'mw:Nowiki' ) |
| 1157 | ) |
| 1158 | ) { |
| 1159 | $i++; |
| 1160 | } |
| 1161 | if ( $i < $n ) { |
| 1162 | // After tokenization, we can get here: |
| 1163 | // * Text explicitly protected by <nowiki> in the parameter. |
| 1164 | // * Other things that should be protected but weren't |
| 1165 | // according to the tokenizer. |
| 1166 | // In template argument, we only need to check for unmatched |
| 1167 | // braces and brackets pairs (which is done in appendStr), |
| 1168 | // but only if they weren't explicitly protected in the |
| 1169 | // passed wikitext. |
| 1170 | $width = $tokens[$i]->dataParsoid->tsr->end - $da->tsr->start; |
| 1171 | $substr = substr( $arg, $da->tsr->start, $width ); |
| 1172 | self::appendStr( |
| 1173 | $substr, |
| 1174 | $last, |
| 1175 | !preg_match( '#<nowiki>[^<]*</nowiki>#', $substr ), |
| 1176 | $buf, |
| 1177 | $openNowiki, |
| 1178 | $isTemplate, |
| 1179 | $serializeAsNamed, |
| 1180 | $opts |
| 1181 | ); |
| 1182 | } |
| 1183 | continue; |
| 1184 | } |
| 1185 | } |
| 1186 | |
| 1187 | switch ( TokenUtils::getTokenType( $t ) ) { |
| 1188 | case 'TagTk': |
| 1189 | case 'EndTagTk': |
| 1190 | case 'NlTk': |
| 1191 | case 'CommentTk': |
| 1192 | $da = $t->dataParsoid; |
| 1193 | if ( empty( $da->tsr ) ) { |
| 1194 | $errors = [ 'Missing tsr for: ' . PHPUtils::jsonEncode( $t ) ]; |
| 1195 | $errors[] = 'Arg : ' . PHPUtils::jsonEncode( $arg ); |
| 1196 | $errors[] = 'Toks: ' . PHPUtils::jsonEncode( $tokens ); |
| 1197 | $env->log( 'error/html2wt/wtescape', implode( "\n", $errors ) ); |
| 1198 | // FIXME $da->tsr will be undefined below. |
| 1199 | // Should we throw an explicit exception here? |
| 1200 | } |
| 1201 | self::appendStr( |
| 1202 | $da->tsr->substr( $arg ), |
| 1203 | $last, |
| 1204 | false, |
| 1205 | $buf, |
| 1206 | $openNowiki, |
| 1207 | $isTemplate, |
| 1208 | $serializeAsNamed, |
| 1209 | $opts |
| 1210 | ); |
| 1211 | break; |
| 1212 | case 'SelfclosingTagTk': |
| 1213 | $da = $t->dataParsoid; |
| 1214 | if ( empty( $da->tsr ) ) { |
| 1215 | $errors = [ 'Missing tsr for: ' . PHPUtils::jsonEncode( $t ) ]; |
| 1216 | $errors[] = 'Arg : ' . PHPUtils::jsonEncode( $arg ); |
| 1217 | $errors[] = 'Toks: ' . PHPUtils::jsonEncode( $tokens ); |
| 1218 | $env->log( 'error/html2wt/wtescape', implode( "\n", $errors ) ); |
| 1219 | // FIXME $da->tsr will be undefined below. |
| 1220 | // Should we throw an explicit exception here? |
| 1221 | } |
| 1222 | $tkSrc = $da->tsr->substr( $arg ); |
| 1223 | // Replace pipe by an entity. This is not completely safe. |
| 1224 | if ( $t->getName() === 'extlink' || $t->getName() === 'urllink' ) { |
| 1225 | $tkBits = $this->tokenizer->tokenizeSync( $tkSrc, [ |
| 1226 | 'startRule' => 'tplarg_or_template_or_bust' |
| 1227 | ] |
| 1228 | ); |
| 1229 | foreach ( $tkBits as $bit ) { |
| 1230 | if ( $bit instanceof Token ) { |
| 1231 | self::appendStr( |
| 1232 | $bit->dataParsoid->src, |
| 1233 | $last, |
| 1234 | false, |
| 1235 | $buf, |
| 1236 | $openNowiki, |
| 1237 | $isTemplate, |
| 1238 | $serializeAsNamed, |
| 1239 | $opts |
| 1240 | ); |
| 1241 | } else { |
| 1242 | // Convert to a named param w/ the same reasoning |
| 1243 | // as above for escapeStr, however, here we replace |
| 1244 | // with an entity to avoid breaking up querystrings |
| 1245 | // with nowikis. |
| 1246 | if ( $isTemplate && !$serializeAsNamed && str_contains( $bit, '=' ) ) { |
| 1247 | if ( $opts['numPositionalArgs'] === 0 |
| 1248 | || $opts['numPositionalArgs'] === $opts['argIndex'] |
| 1249 | ) { |
| 1250 | $serializeAsNamed = true; |
| 1251 | } else { |
| 1252 | $bit = str_replace( '=', '=', $bit ); |
| 1253 | } |
| 1254 | } |
| 1255 | $buf .= str_replace( '|', '|', $bit ); |
| 1256 | } |
| 1257 | } |
| 1258 | } else { |
| 1259 | self::appendStr( |
| 1260 | $tkSrc, |
| 1261 | $last, |
| 1262 | false, |
| 1263 | $buf, |
| 1264 | $openNowiki, |
| 1265 | $isTemplate, |
| 1266 | $serializeAsNamed, |
| 1267 | $opts |
| 1268 | ); |
| 1269 | } |
| 1270 | break; |
| 1271 | case 'string': |
| 1272 | self::appendStr( |
| 1273 | $t, |
| 1274 | $last, |
| 1275 | true, |
| 1276 | $buf, |
| 1277 | $openNowiki, |
| 1278 | $isTemplate, |
| 1279 | $serializeAsNamed, |
| 1280 | $opts |
| 1281 | ); |
| 1282 | break; |
| 1283 | case 'EOFTk': |
| 1284 | break; |
| 1285 | } |
| 1286 | } |
| 1287 | |
| 1288 | // If nowiki still open, close it now. |
| 1289 | if ( $openNowiki ) { |
| 1290 | $buf .= '</nowiki>'; |
| 1291 | } |
| 1292 | |
| 1293 | return [ 'serializeAsNamed' => $serializeAsNamed, 'v' => $buf ]; |
| 1294 | } |
| 1295 | |
| 1296 | /** |
| 1297 | * See also `escapeLinkTarget` in LinkHandler.php |
| 1298 | * |
| 1299 | * @param SerializerState $state |
| 1300 | * @param string $str |
| 1301 | * @param bool $solState |
| 1302 | * @param Node $node |
| 1303 | * @param bool $isMedia |
| 1304 | * @return string |
| 1305 | */ |
| 1306 | public function escapeLinkContent( |
| 1307 | SerializerState $state, string $str, bool $solState, Node $node, bool $isMedia |
| 1308 | ): string { |
| 1309 | // Entity-escape the content. |
| 1310 | $str = Utils::escapeWtEntities( $str ); |
| 1311 | |
| 1312 | // Wikitext-escape content. |
| 1313 | $state->onSOL = $solState; |
| 1314 | $state->wteHandlerStack[] = $isMedia |
| 1315 | ? [ $this, 'mediaOptionHandler' ] |
| 1316 | : [ $this, 'wikilinkHandler' ]; |
| 1317 | $state->inLink = true; |
| 1318 | $res = $this->escapeWikitext( $state, $str, [ 'node' => $node ] ); |
| 1319 | $state->inLink = false; |
| 1320 | array_pop( $state->wteHandlerStack ); |
| 1321 | |
| 1322 | return $res; |
| 1323 | } |
| 1324 | } |