includes/parser/Preprocessor_Hash.php

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