includes/parser/Preprocessor_DOM.php

   1 <?php
   2 /**
   3  * Preprocessor using PHP's dom extension
   4  *
   5  * This program is free software; you can redistribute it and/or modify
   6  * it under the terms of the GNU General Public License as published by
   7  * the Free Software Foundation; either version 2 of the License, or
   8  * (at your option) any later version.
   9  *
  10  * This program is distributed in the hope that it will be useful,
  11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  13  * GNU General Public License for more details.
  14  *
  15  * You should have received a copy of the GNU General Public License along
  16  * with this program; if not, write to the Free Software Foundation, Inc.,
  17  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  18  * http://www.gnu.org/copyleft/gpl.html
  19  *
  20  * @file
  21  * @ingroup Parser
  22  */
  23
  24 /**
  25  * @ingroup Parser
  26  */
  27 // phpcs:ignore Squiz.Classes.ValidClassName.NotCamelCaps
  28 class Preprocessor_DOM extends Preprocessor {
  29
  30         /**
  31          * @var Parser
  32          */
  33         public $parser;
  34
  35         public $memoryLimit;
  36
  37         const CACHE_PREFIX = 'preprocess-xml';
  38
  39         public function __construct( $parser ) {
  40                 $this->parser = $parser;
  41                 $mem = ini_get( 'memory_limit' );
  42                 $this->memoryLimit = false;
  43                 if ( strval( $mem ) !== '' && $mem != -1 ) {
  44                         if ( preg_match( '/^\d+$/', $mem ) ) {
  45                                 $this->memoryLimit = $mem;
  46                         } elseif ( preg_match( '/^(\d+)M$/i', $mem, $m ) ) {
  47                                 $this->memoryLimit = $m[1] * 1048576;
  48                         }
  49                 }
  50         }
  51
  52         /**
  53          * @return PPFrame_DOM
  54          */
  55         public function newFrame() {
  56                 return new PPFrame_DOM( $this );
  57         }
  58
  59         /**
  60          * @param array $args
  61          * @return PPCustomFrame_DOM
  62          */
  63         public function newCustomFrame( $args ) {
  64                 return new PPCustomFrame_DOM( $this, $args );
  65         }
  66
  67         /**
  68          * @param array $values
  69          * @return PPNode_DOM
  70          * @throws MWException
  71          */
  72         public function newPartNodeArray( $values ) {
  73                 // NOTE: DOM manipulation is slower than building & parsing XML! (or so Tim sais)
  74                 $xml = "<list>";
  75
  76                 foreach ( $values as $k => $val ) {
  77                         if ( is_int( $k ) ) {
  78                                 $xml .= "<part><name index=\"$k\"/><value>"
  79                                         . htmlspecialchars( $val ) . "</value></part>";
  80                         } else {
  81                                 $xml .= "<part><name>" . htmlspecialchars( $k )
  82                                         . "</name>=<value>" . htmlspecialchars( $val ) . "</value></part>";
  83                         }
  84                 }
  85
  86                 $xml .= "</list>";
  87
  88                 $dom = new DOMDocument();
  89                 Wikimedia\suppressWarnings();
  90                 $result = $dom->loadXML( $xml );
  91                 Wikimedia\restoreWarnings();
  92                 if ( !$result ) {
  93                         // Try running the XML through UtfNormal to get rid of invalid characters
  94                         $xml = UtfNormal\Validator::cleanUp( $xml );
  95                         // 1 << 19 == XML_PARSE_HUGE, needed so newer versions of libxml2
  96                         // don't barf when the XML is >256 levels deep
  97                         $result = $dom->loadXML( $xml, 1 << 19 );
  98                 }
  99
 100                 if ( !$result ) {
 101                         throw new MWException( 'Parameters passed to ' . __METHOD__ . ' result in invalid XML' );
 102                 }
 103
 104                 $root = $dom->documentElement;
 105                 $node = new PPNode_DOM( $root->childNodes );
 106                 return $node;
 107         }
 108
 109         /**
 110          * @throws MWException
 111          * @return bool
 112          */
 113         public function memCheck() {
 114                 if ( $this->memoryLimit === false ) {
 115                         return true;
 116                 }
 117                 $usage = memory_get_usage();
 118                 if ( $usage > $this->memoryLimit * 0.9 ) {
 119                         $limit = intval( $this->memoryLimit * 0.9 / 1048576 + 0.5 );
 120                         throw new MWException( "Preprocessor hit 90% memory limit ($limit MB)" );
 121                 }
 122                 return $usage <= $this->memoryLimit * 0.8;
 123         }
 124
 125         /**
 126          * Preprocess some wikitext and return the document tree.
 127          * This is the ghost of Parser::replace_variables().
 128          *
 129          * @param string $text The text to parse
 130          * @param int $flags Bitwise combination of:
 131          *     Parser::PTD_FOR_INCLUSION  Handle "<noinclude>" and "<includeonly>"
 132          *                                as if the text is being included. Default
 133          *                                is to assume a direct page view.
 134          *
 135          * The generated DOM tree must depend only on the input text and the flags.
 136          * The DOM tree must be the same in OT_HTML and OT_WIKI mode, to avoid a regression of T6899.
 137          *
 138          * Any flag added to the $flags parameter here, or any other parameter liable to cause a
 139          * change in the DOM tree for a given text, must be passed through the section identifier
 140          * in the section edit link and thus back to extractSections().
 141          *
 142          * The output of this function is currently only cached in process memory, but a persistent
 143          * cache may be implemented at a later date which takes further advantage of these strict
 144          * dependency requirements.
 145          *
 146          * @throws MWException
 147          * @return PPNode_DOM
 148          */
 149         public function preprocessToObj( $text, $flags = 0 ) {
 150                 $xml = $this->cacheGetTree( $text, $flags );
 151                 if ( $xml === false ) {
 152                         $xml = $this->preprocessToXml( $text, $flags );
 153                         $this->cacheSetTree( $text, $flags, $xml );
 154                 }
 155
 156                 // Fail if the number of elements exceeds acceptable limits
 157                 // Do not attempt to generate the DOM
 158                 $this->parser->mGeneratedPPNodeCount += substr_count( $xml, '<' );
 159                 $max = $this->parser->mOptions->getMaxGeneratedPPNodeCount();
 160                 if ( $this->parser->mGeneratedPPNodeCount > $max ) {
 161                         // if ( $cacheable ) { ... }
 162                         throw new MWException( __METHOD__ . ': generated node count limit exceeded' );
 163                 }
 164
 165                 $dom = new DOMDocument;
 166                 Wikimedia\suppressWarnings();
 167                 $result = $dom->loadXML( $xml );
 168                 Wikimedia\restoreWarnings();
 169                 if ( !$result ) {
 170                         // Try running the XML through UtfNormal to get rid of invalid characters
 171                         $xml = UtfNormal\Validator::cleanUp( $xml );
 172                         // 1 << 19 == XML_PARSE_HUGE, needed so newer versions of libxml2
 173                         // don't barf when the XML is >256 levels deep.
 174                         $result = $dom->loadXML( $xml, 1 << 19 );
 175                 }
 176                 if ( $result ) {
 177                         $obj = new PPNode_DOM( $dom->documentElement );
 178                 }
 179
 180                 // if ( $cacheable ) { ... }
 181
 182                 if ( !$result ) {
 183                         throw new MWException( __METHOD__ . ' generated invalid XML' );
 184                 }
 185                 return $obj;
 186         }
 187
 188         /**
 189          * @param string $text
 190          * @param int $flags
 191          * @return string
 192          */
 193         public function preprocessToXml( $text, $flags = 0 ) {
 194                 global $wgDisableLangConversion;
 195
 196                 $forInclusion = $flags & Parser::PTD_FOR_INCLUSION;
 197
 198                 $xmlishElements = $this->parser->getStripList();
 199                 $xmlishAllowMissingEndTag = [ 'includeonly', 'noinclude', 'onlyinclude' ];
 200                 $enableOnlyinclude = false;
 201                 if ( $forInclusion ) {
 202                         $ignoredTags = [ 'includeonly', '/includeonly' ];
 203                         $ignoredElements = [ 'noinclude' ];
 204                         $xmlishElements[] = 'noinclude';
 205                         if ( strpos( $text, '<onlyinclude>' ) !== false
 206                                 && strpos( $text, '</onlyinclude>' ) !== false
 207                         ) {
 208                                 $enableOnlyinclude = true;
 209                         }
 210                 } else {
 211                         $ignoredTags = [ 'noinclude', '/noinclude', 'onlyinclude', '/onlyinclude' ];
 212                         $ignoredElements = [ 'includeonly' ];
 213                         $xmlishElements[] = 'includeonly';
 214                 }
 215                 $xmlishRegex = implode( '|', array_merge( $xmlishElements, $ignoredTags ) );
 216
 217                 // Use "A" modifier (anchored) instead of "^", because ^ doesn't work with an offset
 218                 $elementsRegex = "~($xmlishRegex)(?:\s|\/>|>)|(!--)~iA";
 219
 220                 $stack = new PPDStack;
 221
 222                 $searchBase = "[{<\n"; # }
 223                 if ( !$wgDisableLangConversion ) {
 224                         $searchBase .= '-';
 225                 }
 226
 227                 // For fast reverse searches
 228                 $revText = strrev( $text );
 229                 $lengthText = strlen( $text );
 230
 231                 // Input pointer, starts out pointing to a pseudo-newline before the start
 232                 $i = 0;
 233                 // Current accumulator
 234                 $accum =& $stack->getAccum();
 235                 $accum = '<root>';
 236                 // True to find equals signs in arguments
 237                 $findEquals = false;
 238                 // True to take notice of pipe characters
 239                 $findPipe = false;
 240                 $headingIndex = 1;
 241                 // True if $i is inside a possible heading
 242                 $inHeading = false;
 243                 // True if there are no more greater-than (>) signs right of $i
 244                 $noMoreGT = false;
 245                 // Map of tag name => true if there are no more closing tags of given type right of $i
 246                 $noMoreClosingTag = [];
 247                 // True to ignore all input up to the next <onlyinclude>
 248                 $findOnlyinclude = $enableOnlyinclude;
 249                 // Do a line-start run without outputting an LF character
 250                 $fakeLineStart = true;
 251
 252                 while ( true ) {
 253                         // $this->memCheck();
 254
 255                         if ( $findOnlyinclude ) {
 256                                 // Ignore all input up to the next <onlyinclude>
 257                                 $startPos = strpos( $text, '<onlyinclude>', $i );
 258                                 if ( $startPos === false ) {
 259                                         // Ignored section runs to the end
 260                                         $accum .= '<ignore>' . htmlspecialchars( substr( $text, $i ) ) . '</ignore>';
 261                                         break;
 262                                 }
 263                                 $tagEndPos = $startPos + strlen( '<onlyinclude>' ); // past-the-end
 264                                 $accum .= '<ignore>' . htmlspecialchars( substr( $text, $i, $tagEndPos - $i ) ) . '</ignore>';
 265                                 $i = $tagEndPos;
 266                                 $findOnlyinclude = false;
 267                         }
 268
 269                         if ( $fakeLineStart ) {
 270                                 $found = 'line-start';
 271                                 $curChar = '';
 272                         } else {
 273                                 # Find next opening brace, closing brace or pipe
 274                                 $search = $searchBase;
 275                                 if ( $stack->top === false ) {
 276                                         $currentClosing = '';
 277                                 } else {
 278                                         $currentClosing = $stack->top->close;
 279                                         $search .= $currentClosing;
 280                                 }
 281                                 if ( $findPipe ) {
 282                                         $search .= '|';
 283                                 }
 284                                 if ( $findEquals ) {
 285                                         // First equals will be for the template
 286                                         $search .= '=';
 287                                 }
 288                                 $rule = null;
 289                                 # Output literal section, advance input counter
 290                                 $literalLength = strcspn( $text, $search, $i );
 291                                 if ( $literalLength > 0 ) {
 292                                         $accum .= htmlspecialchars( substr( $text, $i, $literalLength ) );
 293                                         $i += $literalLength;
 294                                 }
 295                                 if ( $i >= $lengthText ) {
 296                                         if ( $currentClosing == "\n" ) {
 297                                                 // Do a past-the-end run to finish off the heading
 298                                                 $curChar = '';
 299                                                 $found = 'line-end';
 300                                         } else {
 301                                                 # All done
 302                                                 break;
 303                                         }
 304                                 } else {
 305                                         $curChar = $curTwoChar = $text[$i];
 306                                         if ( ( $i + 1 ) < $lengthText ) {
 307                                                 $curTwoChar .= $text[$i + 1];
 308                                         }
 309                                         if ( $curChar == '|' ) {
 310                                                 $found = 'pipe';
 311                                         } elseif ( $curChar == '=' ) {
 312                                                 $found = 'equals';
 313                                         } elseif ( $curChar == '<' ) {
 314                                                 $found = 'angle';
 315                                         } elseif ( $curChar == "\n" ) {
 316                                                 if ( $inHeading ) {
 317                                                         $found = 'line-end';
 318                                                 } else {
 319                                                         $found = 'line-start';
 320                                                 }
 321                                         } elseif ( $curTwoChar == $currentClosing ) {
 322                                                 $found = 'close';
 323                                                 $curChar = $curTwoChar;
 324                                         } elseif ( $curChar == $currentClosing ) {
 325                                                 $found = 'close';
 326                                         } elseif ( isset( $this->rules[$curTwoChar] ) ) {
 327                                                 $curChar = $curTwoChar;
 328                                                 $found = 'open';
 329                                                 $rule = $this->rules[$curChar];
 330                                         } elseif ( isset( $this->rules[$curChar] ) ) {
 331                                                 $found = 'open';
 332                                                 $rule = $this->rules[$curChar];
 333                                         } else {
 334                                                 # Some versions of PHP have a strcspn which stops on
 335                                                 # null characters; ignore these and continue.
 336                                                 # We also may get '-' and '}' characters here which
 337                                                 # don't match -{ or $currentClosing.  Add these to
 338                                                 # output and continue.
 339                                                 if ( $curChar == '-' || $curChar == '}' ) {
 340                                                         $accum .= $curChar;
 341                                                 }
 342                                                 ++$i;
 343                                                 continue;
 344                                         }
 345                                 }
 346                         }
 347
 348                         if ( $found == 'angle' ) {
 349                                 $matches = false;
 350                                 // Handle </onlyinclude>
 351                                 if ( $enableOnlyinclude
 352                                         && substr( $text, $i, strlen( '</onlyinclude>' ) ) == '</onlyinclude>'
 353                                 ) {
 354                                         $findOnlyinclude = true;
 355                                         continue;
 356                                 }
 357
 358                                 // Determine element name
 359                                 if ( !preg_match( $elementsRegex, $text, $matches, 0, $i + 1 ) ) {
 360                                         // Element name missing or not listed
 361                                         $accum .= '&lt;';
 362                                         ++$i;
 363                                         continue;
 364                                 }
 365                                 // Handle comments
 366                                 if ( isset( $matches[2] ) && $matches[2] == '!--' ) {
 367                                         // To avoid leaving blank lines, when a sequence of
 368                                         // space-separated comments is both preceded and followed by
 369                                         // a newline (ignoring spaces), then
 370                                         // trim leading and trailing spaces and the trailing newline.
 371
 372                                         // Find the end
 373                                         $endPos = strpos( $text, '-->', $i + 4 );
 374                                         if ( $endPos === false ) {
 375                                                 // Unclosed comment in input, runs to end
 376                                                 $inner = substr( $text, $i );
 377                                                 $accum .= '<comment>' . htmlspecialchars( $inner ) . '</comment>';
 378                                                 $i = $lengthText;
 379                                         } else {
 380                                                 // Search backwards for leading whitespace
 381                                                 $wsStart = $i ? ( $i - strspn( $revText, " \t", $lengthText - $i ) ) : 0;
 382
 383                                                 // Search forwards for trailing whitespace
 384                                                 // $wsEnd will be the position of the last space (or the '>' if there's none)
 385                                                 $wsEnd = $endPos + 2 + strspn( $text, " \t", $endPos + 3 );
 386
 387                                                 // Keep looking forward as long as we're finding more
 388                                                 // comments.
 389                                                 $comments = [ [ $wsStart, $wsEnd ] ];
 390                                                 while ( substr( $text, $wsEnd + 1, 4 ) == '<!--' ) {
 391                                                         $c = strpos( $text, '-->', $wsEnd + 4 );
 392                                                         if ( $c === false ) {
 393                                                                 break;
 394                                                         }
 395                                                         $c = $c + 2 + strspn( $text, " \t", $c + 3 );
 396                                                         $comments[] = [ $wsEnd + 1, $c ];
 397                                                         $wsEnd = $c;
 398                                                 }
 399
 400                                                 // Eat the line if possible
 401                                                 // TODO: This could theoretically be done if $wsStart == 0, i.e. for comments at
 402                                                 // the overall start. That's not how Sanitizer::removeHTMLcomments() did it, but
 403                                                 // it's a possible beneficial b/c break.
 404                                                 if ( $wsStart > 0 && substr( $text, $wsStart - 1, 1 ) == "\n"
 405                                                         && substr( $text, $wsEnd + 1, 1 ) == "\n"
 406                                                 ) {
 407                                                         // Remove leading whitespace from the end of the accumulator
 408                                                         // Sanity check first though
 409                                                         $wsLength = $i - $wsStart;
 410                                                         if ( $wsLength > 0
 411                                                                 && strspn( $accum, " \t", -$wsLength ) === $wsLength
 412                                                         ) {
 413                                                                 $accum = substr( $accum, 0, -$wsLength );
 414                                                         }
 415
 416                                                         // Dump all but the last comment to the accumulator
 417                                                         foreach ( $comments as $j => $com ) {
 418                                                                 $startPos = $com[0];
 419                                                                 $endPos = $com[1] + 1;
 420                                                                 if ( $j == ( count( $comments ) - 1 ) ) {
 421                                                                         break;
 422                                                                 }
 423                                                                 $inner = substr( $text, $startPos, $endPos - $startPos );
 424                                                                 $accum .= '<comment>' . htmlspecialchars( $inner ) . '</comment>';
 425                                                         }
 426
 427                                                         // Do a line-start run next time to look for headings after the comment
 428                                                         $fakeLineStart = true;
 429                                                 } else {
 430                                                         // No line to eat, just take the comment itself
 431                                                         $startPos = $i;
 432                                                         $endPos += 2;
 433                                                 }
 434
 435                                                 if ( $stack->top ) {
 436                                                         $part = $stack->top->getCurrentPart();
 437                                                         if ( !( isset( $part->commentEnd ) && $part->commentEnd == $wsStart - 1 ) ) {
 438                                                                 $part->visualEnd = $wsStart;
 439                                                         }
 440                                                         // Else comments abutting, no change in visual end
 441                                                         $part->commentEnd = $endPos;
 442                                                 }
 443                                                 $i = $endPos + 1;
 444                                                 $inner = substr( $text, $startPos, $endPos - $startPos + 1 );
 445                                                 $accum .= '<comment>' . htmlspecialchars( $inner ) . '</comment>';
 446                                         }
 447                                         continue;
 448                                 }
 449                                 $name = $matches[1];
 450                                 $lowerName = strtolower( $name );
 451                                 $attrStart = $i + strlen( $name ) + 1;
 452
 453                                 // Find end of tag
 454                                 $tagEndPos = $noMoreGT ? false : strpos( $text, '>', $attrStart );
 455                                 if ( $tagEndPos === false ) {
 456                                         // Infinite backtrack
 457                                         // Disable tag search to prevent worst-case O(N^2) performance
 458                                         $noMoreGT = true;
 459                                         $accum .= '&lt;';
 460                                         ++$i;
 461                                         continue;
 462                                 }
 463
 464                                 // Handle ignored tags
 465                                 if ( in_array( $lowerName, $ignoredTags ) ) {
 466                                         $accum .= '<ignore>'
 467                                                 . htmlspecialchars( substr( $text, $i, $tagEndPos - $i + 1 ) )
 468                                                 . '</ignore>';
 469                                         $i = $tagEndPos + 1;
 470                                         continue;
 471                                 }
 472
 473                                 $tagStartPos = $i;
 474                                 if ( $text[$tagEndPos - 1] == '/' ) {
 475                                         $attrEnd = $tagEndPos - 1;
 476                                         $inner = null;
 477                                         $i = $tagEndPos + 1;
 478                                         $close = '';
 479                                 } else {
 480                                         $attrEnd = $tagEndPos;
 481                                         // Find closing tag
 482                                         if (
 483                                                 !isset( $noMoreClosingTag[$name] ) &&
 484                                                 preg_match( "/<\/" . preg_quote( $name, '/' ) . "\s*>/i",
 485                                                         $text, $matches, PREG_OFFSET_CAPTURE, $tagEndPos + 1 )
 486                                         ) {
 487                                                 $inner = substr( $text, $tagEndPos + 1, $matches[0][1] - $tagEndPos - 1 );
 488                                                 $i = $matches[0][1] + strlen( $matches[0][0] );
 489                                                 $close = '<close>' . htmlspecialchars( $matches[0][0] ) . '</close>';
 490                                         } else {
 491                                                 // No end tag
 492                                                 if ( in_array( $name, $xmlishAllowMissingEndTag ) ) {
 493                                                         // Let it run out to the end of the text.
 494                                                         $inner = substr( $text, $tagEndPos + 1 );
 495                                                         $i = $lengthText;
 496                                                         $close = '';
 497                                                 } else {
 498                                                         // Don't match the tag, treat opening tag as literal and resume parsing.
 499                                                         $i = $tagEndPos + 1;
 500                                                         $accum .= htmlspecialchars( substr( $text, $tagStartPos, $tagEndPos + 1 - $tagStartPos ) );
 501                                                         // Cache results, otherwise we have O(N^2) performance for input like <foo><foo><foo>...
 502                                                         $noMoreClosingTag[$name] = true;
 503                                                         continue;
 504                                                 }
 505                                         }
 506                                 }
 507                                 // <includeonly> and <noinclude> just become <ignore> tags
 508                                 if ( in_array( $lowerName, $ignoredElements ) ) {
 509                                         $accum .= '<ignore>' . htmlspecialchars( substr( $text, $tagStartPos, $i - $tagStartPos ) )
 510                                                 . '</ignore>';
 511                                         continue;
 512                                 }
 513
 514                                 $accum .= '<ext>';
 515                                 if ( $attrEnd <= $attrStart ) {
 516                                         $attr = '';
 517                                 } else {
 518                                         $attr = substr( $text, $attrStart, $attrEnd - $attrStart );
 519                                 }
 520                                 $accum .= '<name>' . htmlspecialchars( $name ) . '</name>' .
 521                                         // Note that the attr element contains the whitespace between name and attribute,
 522                                         // this is necessary for precise reconstruction during pre-save transform.
 523                                         '<attr>' . htmlspecialchars( $attr ) . '</attr>';
 524                                 if ( $inner !== null ) {
 525                                         $accum .= '<inner>' . htmlspecialchars( $inner ) . '</inner>';
 526                                 }
 527                                 $accum .= $close . '</ext>';
 528                         } elseif ( $found == 'line-start' ) {
 529                                 // Is this the start of a heading?
 530                                 // Line break belongs before the heading element in any case
 531                                 if ( $fakeLineStart ) {
 532                                         $fakeLineStart = false;
 533                                 } else {
 534                                         $accum .= $curChar;
 535                                         $i++;
 536                                 }
 537
 538                                 $count = strspn( $text, '=', $i, 6 );
 539                                 if ( $count == 1 && $findEquals ) {
 540                                         // DWIM: This looks kind of like a name/value separator.
 541                                         // Let's let the equals handler have it and break the
 542                                         // potential heading. This is heuristic, but AFAICT the
 543                                         // methods for completely correct disambiguation are very
 544                                         // complex.
 545                                 } elseif ( $count > 0 ) {
 546                                         $piece = [
 547                                                 'open' => "\n",
 548                                                 'close' => "\n",
 549                                                 'parts' => [ new PPDPart( str_repeat( '=', $count ) ) ],
 550                                                 'startPos' => $i,
 551                                                 'count' => $count ];
 552                                         $stack->push( $piece );
 553                                         $accum =& $stack->getAccum();
 554                                         $stackFlags = $stack->getFlags();
 555                                         if ( isset( $stackFlags['findEquals'] ) ) {
 556                                                 $findEquals = $stackFlags['findEquals'];
 557                                         }
 558                                         if ( isset( $stackFlags['findPipe'] ) ) {
 559                                                 $findPipe = $stackFlags['findPipe'];
 560                                         }
 561                                         if ( isset( $stackFlags['inHeading'] ) ) {
 562                                                 $inHeading = $stackFlags['inHeading'];
 563                                         }
 564                                         $i += $count;
 565                                 }
 566                         } elseif ( $found == 'line-end' ) {
 567                                 $piece = $stack->top;
 568                                 // A heading must be open, otherwise \n wouldn't have been in the search list
 569                                 // FIXME: Don't use assert()
 570                                 // phpcs:ignore MediaWiki.Usage.ForbiddenFunctions.assert
 571                                 assert( $piece->open === "\n" );
 572                                 $part = $piece->getCurrentPart();
 573                                 // Search back through the input to see if it has a proper close.
 574                                 // Do this using the reversed string since the other solutions
 575                                 // (end anchor, etc.) are inefficient.
 576                                 $wsLength = strspn( $revText, " \t", $lengthText - $i );
 577                                 $searchStart = $i - $wsLength;
 578                                 if ( isset( $part->commentEnd ) && $searchStart - 1 == $part->commentEnd ) {
 579                                         // Comment found at line end
 580                                         // Search for equals signs before the comment
 581                                         $searchStart = $part->visualEnd;
 582                                         $searchStart -= strspn( $revText, " \t", $lengthText - $searchStart );
 583                                 }
 584                                 $count = $piece->count;
 585                                 $equalsLength = strspn( $revText, '=', $lengthText - $searchStart );
 586                                 if ( $equalsLength > 0 ) {
 587                                         if ( $searchStart - $equalsLength == $piece->startPos ) {
 588                                                 // This is just a single string of equals signs on its own line
 589                                                 // Replicate the doHeadings behavior /={count}(.+)={count}/
 590                                                 // First find out how many equals signs there really are (don't stop at 6)
 591                                                 $count = $equalsLength;
 592                                                 if ( $count < 3 ) {
 593                                                         $count = 0;
 594                                                 } else {
 595                                                         $count = min( 6, intval( ( $count - 1 ) / 2 ) );
 596                                                 }
 597                                         } else {
 598                                                 $count = min( $equalsLength, $count );
 599                                         }
 600                                         if ( $count > 0 ) {
 601                                                 // Normal match, output <h>
 602                                                 $element = "<h level=\"$count\" i=\"$headingIndex\">$accum</h>";
 603                                                 $headingIndex++;
 604                                         } else {
 605                                                 // Single equals sign on its own line, count=0
 606                                                 $element = $accum;
 607                                         }
 608                                 } else {
 609                                         // No match, no <h>, just pass down the inner text
 610                                         $element = $accum;
 611                                 }
 612                                 // Unwind the stack
 613                                 $stack->pop();
 614                                 $accum =& $stack->getAccum();
 615                                 $stackFlags = $stack->getFlags();
 616                                 if ( isset( $stackFlags['findEquals'] ) ) {
 617                                         $findEquals = $stackFlags['findEquals'];
 618                                 }
 619                                 if ( isset( $stackFlags['findPipe'] ) ) {
 620                                         $findPipe = $stackFlags['findPipe'];
 621                                 }
 622                                 if ( isset( $stackFlags['inHeading'] ) ) {
 623                                         $inHeading = $stackFlags['inHeading'];
 624                                 }
 625
 626                                 // Append the result to the enclosing accumulator
 627                                 $accum .= $element;
 628                                 // Note that we do NOT increment the input pointer.
 629                                 // This is because the closing linebreak could be the opening linebreak of
 630                                 // another heading. Infinite loops are avoided because the next iteration MUST
 631                                 // hit the heading open case above, which unconditionally increments the
 632                                 // input pointer.
 633                         } elseif ( $found == 'open' ) {
 634                                 # count opening brace characters
 635                                 $curLen = strlen( $curChar );
 636                                 $count = ( $curLen > 1 ) ?
 637                                         # allow the final character to repeat
 638                                         strspn( $text, $curChar[$curLen - 1], $i + 1 ) + 1 :
 639                                         strspn( $text, $curChar, $i );
 640
 641                                 $savedPrefix = '';
 642                                 $lineStart = ( $i > 0 && $text[$i - 1] == "\n" );
 643
 644                                 if ( $curChar === "-{" && $count > $curLen ) {
 645                                         // -{ => {{ transition because rightmost wins
 646                                         $savedPrefix = '-';
 647                                         $i++;
 648                                         $curChar = '{';
 649                                         $count--;
 650                                         $rule = $this->rules[$curChar];
 651                                 }
 652
 653                                 # we need to add to stack only if opening brace count is enough for one of the rules
 654                                 if ( $count >= $rule['min'] ) {
 655                                         # Add it to the stack
 656                                         $piece = [
 657                                                 'open' => $curChar,
 658                                                 'close' => $rule['end'],
 659                                                 'savedPrefix' => $savedPrefix,
 660                                                 'count' => $count,
 661                                                 'lineStart' => $lineStart,
 662                                         ];
 663
 664                                         $stack->push( $piece );
 665                                         $accum =& $stack->getAccum();
 666                                         $stackFlags = $stack->getFlags();
 667                                         if ( isset( $stackFlags['findEquals'] ) ) {
 668                                                 $findEquals = $stackFlags['findEquals'];
 669                                         }
 670                                         if ( isset( $stackFlags['findPipe'] ) ) {
 671                                                 $findPipe = $stackFlags['findPipe'];
 672                                         }
 673                                         if ( isset( $stackFlags['inHeading'] ) ) {
 674                                                 $inHeading = $stackFlags['inHeading'];
 675                                         }
 676                                 } else {
 677                                         # Add literal brace(s)
 678                                         $accum .= htmlspecialchars( $savedPrefix . str_repeat( $curChar, $count ) );
 679                                 }
 680                                 $i += $count;
 681                         } elseif ( $found == 'close' ) {
 682                                 $piece = $stack->top;
 683                                 # lets check if there are enough characters for closing brace
 684                                 $maxCount = $piece->count;
 685                                 if ( $piece->close === '}-' && $curChar === '}' ) {
 686                                         $maxCount--; # don't try to match closing '-' as a '}'
 687                                 }
 688                                 $curLen = strlen( $curChar );
 689                                 $count = ( $curLen > 1 ) ? $curLen :
 690                                         strspn( $text, $curChar, $i, $maxCount );
 691
 692                                 # check for maximum matching characters (if there are 5 closing
 693                                 # characters, we will probably need only 3 - depending on the rules)
 694                                 $rule = $this->rules[$piece->open];
 695                                 if ( $count > $rule['max'] ) {
 696                                         # The specified maximum exists in the callback array, unless the caller
 697                                         # has made an error
 698                                         $matchingCount = $rule['max'];
 699                                 } else {
 700                                         # Count is less than the maximum
 701                                         # Skip any gaps in the callback array to find the true largest match
 702                                         # Need to use array_key_exists not isset because the callback can be null
 703                                         $matchingCount = $count;
 704                                         while ( $matchingCount > 0 && !array_key_exists( $matchingCount, $rule['names'] ) ) {
 705                                                 --$matchingCount;
 706                                         }
 707                                 }
 708
 709                                 if ( $matchingCount <= 0 ) {
 710                                         # No matching element found in callback array
 711                                         # Output a literal closing brace and continue
 712                                         $endText = substr( $text, $i, $count );
 713                                         $accum .= htmlspecialchars( $endText );
 714                                         $i += $count;
 715                                         continue;
 716                                 }
 717                                 $name = $rule['names'][$matchingCount];
 718                                 if ( $name === null ) {
 719                                         // No element, just literal text
 720                                         $endText = substr( $text, $i, $matchingCount );
 721                                         $element = $piece->breakSyntax( $matchingCount ) . $endText;
 722                                 } else {
 723                                         # Create XML element
 724                                         # Note: $parts is already XML, does not need to be encoded further
 725                                         $parts = $piece->parts;
 726                                         $title = $parts[0]->out;
 727                                         unset( $parts[0] );
 728
 729                                         # The invocation is at the start of the line if lineStart is set in
 730                                         # the stack, and all opening brackets are used up.
 731                                         if ( $maxCount == $matchingCount &&
 732                                                         !empty( $piece->lineStart ) &&
 733                                                         strlen( $piece->savedPrefix ) == 0 ) {
 734                                                 $attr = ' lineStart="1"';
 735                                         } else {
 736                                                 $attr = '';
 737                                         }
 738
 739                                         $element = "<$name$attr>";
 740                                         $element .= "<title>$title</title>";
 741                                         $argIndex = 1;
 742                                         foreach ( $parts as $part ) {
 743                                                 if ( isset( $part->eqpos ) ) {
 744                                                         $argName = substr( $part->out, 0, $part->eqpos );
 745                                                         $argValue = substr( $part->out, $part->eqpos + 1 );
 746                                                         $element .= "<part><name>$argName</name>=<value>$argValue</value></part>";
 747                                                 } else {
 748                                                         $element .= "<part><name index=\"$argIndex\" /><value>{$part->out}</value></part>";
 749                                                         $argIndex++;
 750                                                 }
 751                                         }
 752                                         $element .= "</$name>";
 753                                 }
 754
 755                                 # Advance input pointer
 756                                 $i += $matchingCount;
 757
 758                                 # Unwind the stack
 759                                 $stack->pop();
 760                                 $accum =& $stack->getAccum();
 761
 762                                 # Re-add the old stack element if it still has unmatched opening characters remaining
 763                                 if ( $matchingCount < $piece->count ) {
 764                                         $piece->parts = [ new PPDPart ];
 765                                         $piece->count -= $matchingCount;
 766                                         # do we still qualify for any callback with remaining count?
 767                                         $min = $this->rules[$piece->open]['min'];
 768                                         if ( $piece->count >= $min ) {
 769                                                 $stack->push( $piece );
 770                                                 $accum =& $stack->getAccum();
 771                                         } elseif ( $piece->count == 1 && $piece->open === '{' && $piece->savedPrefix === '-' ) {
 772                                                 $piece->savedPrefix = '';
 773                                                 $piece->open = '-{';
 774                                                 $piece->count = 2;
 775                                                 $piece->close = $this->rules[$piece->open]['end'];
 776                                                 $stack->push( $piece );
 777                                                 $accum =& $stack->getAccum();
 778                                         } else {
 779                                                 $s = substr( $piece->open, 0, -1 );
 780                                                 $s .= str_repeat(
 781                                                         substr( $piece->open, -1 ),
 782                                                         $piece->count - strlen( $s )
 783                                                 );
 784                                                 $accum .= $piece->savedPrefix . $s;
 785                                         }
 786                                 } elseif ( $piece->savedPrefix !== '' ) {
 787                                         $accum .= $piece->savedPrefix;
 788                                 }
 789
 790                                 $stackFlags = $stack->getFlags();
 791                                 if ( isset( $stackFlags['findEquals'] ) ) {
 792                                         $findEquals = $stackFlags['findEquals'];
 793                                 }
 794                                 if ( isset( $stackFlags['findPipe'] ) ) {
 795                                         $findPipe = $stackFlags['findPipe'];
 796                                 }
 797                                 if ( isset( $stackFlags['inHeading'] ) ) {
 798                                         $inHeading = $stackFlags['inHeading'];
 799                                 }
 800
 801                                 # Add XML element to the enclosing accumulator
 802                                 $accum .= $element;
 803                         } elseif ( $found == 'pipe' ) {
 804                                 $findEquals = true; // shortcut for getFlags()
 805                                 $stack->addPart();
 806                                 $accum =& $stack->getAccum();
 807                                 ++$i;
 808                         } elseif ( $found == 'equals' ) {
 809                                 $findEquals = false; // shortcut for getFlags()
 810                                 $stack->getCurrentPart()->eqpos = strlen( $accum );
 811                                 $accum .= '=';
 812                                 ++$i;
 813                         }
 814                 }
 815
 816                 # Output any remaining unclosed brackets
 817                 foreach ( $stack->stack as $piece ) {
 818                         $stack->rootAccum .= $piece->breakSyntax();
 819                 }
 820                 $stack->rootAccum .= '</root>';
 821                 $xml = $stack->rootAccum;
 822
 823                 return $xml;
 824         }
 825 }