/**
* Preprocess some wikitext and return the document tree.
* This is the ghost of Parser::replace_variables().
*
* @param string $text The text to parse
* @param integer flags Bitwise combination of:
* Parser::PTD_FOR_INCLUSION Handle <noinclude>/<includeonly> as if the text is being
* included. Default is to assume a direct page view.
*
* The generated DOM tree must depend only on the input text and the flags.
* The DOM tree must be the same in OT_HTML and OT_WIKI mode, to avoid a regression of bug 4899.
*
* Any flag added to the $flags parameter here, or any other parameter liable to cause a
* change in the DOM tree for a given text, must be passed through the section identifier
* in the section edit link and thus back to extractSections().
*
* The output of this function is currently only cached in process memory, but a persistent
* cache may be implemented at a later date which takes further advantage of these strict
* dependency requirements.
*
* @private
*/
function preprocessToObj($text, $flags = 0)
{
wfProfileIn(__METHOD__);
wfProfileIn(__METHOD__ . '-makexml');
$rules = array('{' => array('end' => '}', 'names' => array(2 => 'template', 3 => 'tplarg'), 'min' => 2, 'max' => 3), '[' => array('end' => ']', 'names' => array(2 => null), 'min' => 2, 'max' => 2));
$forInclusion = $flags & Parser::PTD_FOR_INCLUSION;
$xmlishElements = $this->parser->getStripList();
$enableOnlyinclude = false;
if ($forInclusion) {
$ignoredTags = array('includeonly', '/includeonly');
$ignoredElements = array('noinclude');
$xmlishElements[] = 'noinclude';
if (strpos($text, '<onlyinclude>') !== false && strpos($text, '</onlyinclude>') !== false) {
$enableOnlyinclude = true;
}
} else {
$ignoredTags = array('noinclude', '/noinclude', 'onlyinclude', '/onlyinclude');
$ignoredElements = array('includeonly');
$xmlishElements[] = 'includeonly';
}
$xmlishRegex = implode('|', array_merge($xmlishElements, $ignoredTags));
// Use "A" modifier (anchored) instead of "^", because ^ doesn't work with an offset
$elementsRegex = "~({$xmlishRegex})(?:\\s|\\/>|>)|(!--)~iA";
$stack = new PPDStack();
$searchBase = "[{<\n";
#}
$revText = strrev($text);
// For fast reverse searches
$i = 0;
# Input pointer, starts out pointing to a pseudo-newline before the start
$accum =& $stack->getAccum();
# Current accumulator
$accum = '<root>';
$findEquals = false;
# True to find equals signs in arguments
$findPipe = false;
# True to take notice of pipe characters
$headingIndex = 1;
$inHeading = false;
# True if $i is inside a possible heading
$noMoreGT = false;
# True if there are no more greater-than (>) signs right of $i
$findOnlyinclude = $enableOnlyinclude;
# True to ignore all input up to the next <onlyinclude>
$fakeLineStart = true;
# Do a line-start run without outputting an LF character
while (true) {
//$this->memCheck();
if ($findOnlyinclude) {
// Ignore all input up to the next <onlyinclude>
$startPos = strpos($text, '<onlyinclude>', $i);
if ($startPos === false) {
// Ignored section runs to the end
$accum .= '<ignore>' . htmlspecialchars(substr($text, $i)) . '</ignore>';
break;
}
$tagEndPos = $startPos + strlen('<onlyinclude>');
// past-the-end
$accum .= '<ignore>' . htmlspecialchars(substr($text, $i, $tagEndPos - $i)) . '</ignore>';
$i = $tagEndPos;
$findOnlyinclude = false;
}
if ($fakeLineStart) {
$found = 'line-start';
$curChar = '';
} else {
# Find next opening brace, closing brace or pipe
$search = $searchBase;
if ($stack->top === false) {
$currentClosing = '';
} else {
$currentClosing = $stack->top->close;
$search .= $currentClosing;
}
if ($findPipe) {
$search .= '|';
}
if ($findEquals) {
// First equals will be for the template
$search .= '=';
}
$rule = null;
# Output literal section, advance input counter
$literalLength = strcspn($text, $search, $i);
if ($literalLength > 0) {
$accum .= htmlspecialchars(substr($text, $i, $literalLength));
$i += $literalLength;
}
if ($i >= strlen($text)) {
if ($currentClosing == "\n") {
// Do a past-the-end run to finish off the heading
$curChar = '';
$found = 'line-end';
} else {
# All done
break;
}
} else {
$curChar = $text[$i];
if ($curChar == '|') {
$found = 'pipe';
} elseif ($curChar == '=') {
$found = 'equals';
} elseif ($curChar == '<') {
$found = 'angle';
} elseif ($curChar == "\n") {
if ($inHeading) {
$found = 'line-end';
} else {
$found = 'line-start';
}
} elseif ($curChar == $currentClosing) {
$found = 'close';
} elseif (isset($rules[$curChar])) {
$found = 'open';
$rule = $rules[$curChar];
} else {
# Some versions of PHP have a strcspn which stops on null characters
# Ignore and continue
++$i;
continue;
}
}
}
if ($found == 'angle') {
$matches = false;
// Handle </onlyinclude>
if ($enableOnlyinclude && substr($text, $i, strlen('</onlyinclude>')) == '</onlyinclude>') {
$findOnlyinclude = true;
continue;
}
// Determine element name
if (!preg_match($elementsRegex, $text, $matches, 0, $i + 1)) {
// Element name missing or not listed
$accum .= '<';
++$i;
continue;
}
// Handle comments
if (isset($matches[2]) && $matches[2] == '!--') {
// To avoid leaving blank lines, when a comment is both preceded
// and followed by a newline (ignoring spaces), trim leading and
// trailing spaces and one of the newlines.
// Find the end
$endPos = strpos($text, '-->', $i + 4);
if ($endPos === false) {
// Unclosed comment in input, runs to end
$inner = substr($text, $i);
$accum .= '<comment>' . htmlspecialchars($inner) . '</comment>';
$i = strlen($text);
} else {
// Search backwards for leading whitespace
$wsStart = $i ? $i - strspn($revText, ' ', strlen($text) - $i) : 0;
// Search forwards for trailing whitespace
// $wsEnd will be the position of the last space
$wsEnd = $endPos + 2 + strspn($text, ' ', $endPos + 3);
// Eat the line if possible
// TODO: This could theoretically be done if $wsStart == 0, i.e. for comments at
// the overall start. That's not how Sanitizer::removeHTMLcomments() did it, but
// it's a possible beneficial b/c break.
if ($wsStart > 0 && substr($text, $wsStart - 1, 1) == "\n" && substr($text, $wsEnd + 1, 1) == "\n") {
$startPos = $wsStart;
$endPos = $wsEnd + 1;
// Remove leading whitespace from the end of the accumulator
// Sanity check first though
$wsLength = $i - $wsStart;
if ($wsLength > 0 && substr($accum, -$wsLength) === str_repeat(' ', $wsLength)) {
$accum = substr($accum, 0, -$wsLength);
}
// Do a line-start run next time to look for headings after the comment
$fakeLineStart = true;
} else {
// No line to eat, just take the comment itself
$startPos = $i;
$endPos += 2;
}
if ($stack->top) {
$part = $stack->top->getCurrentPart();
if (isset($part->commentEnd) && $part->commentEnd == $wsStart - 1) {
// Comments abutting, no change in visual end
$part->commentEnd = $wsEnd;
} else {
$part->visualEnd = $wsStart;
$part->commentEnd = $endPos;
}
}
$i = $endPos + 1;
$inner = substr($text, $startPos, $endPos - $startPos + 1);
$accum .= '<comment>' . htmlspecialchars($inner) . '</comment>';
}
continue;
}
$name = $matches[1];
$lowerName = strtolower($name);
$attrStart = $i + strlen($name) + 1;
// Find end of tag
$tagEndPos = $noMoreGT ? false : strpos($text, '>', $attrStart);
if ($tagEndPos === false) {
// Infinite backtrack
// Disable tag search to prevent worst-case O(N^2) performance
$noMoreGT = true;
$accum .= '<';
++$i;
continue;
}
// Handle ignored tags
if (in_array($lowerName, $ignoredTags)) {
$accum .= '<ignore>' . htmlspecialchars(substr($text, $i, $tagEndPos - $i + 1)) . '</ignore>';
$i = $tagEndPos + 1;
continue;
}
$tagStartPos = $i;
if ($text[$tagEndPos - 1] == '/') {
$attrEnd = $tagEndPos - 1;
$inner = null;
$i = $tagEndPos + 1;
$close = '';
} else {
$attrEnd = $tagEndPos;
// Find closing tag
if (preg_match("/<\\/{$name}\\s*>/i", $text, $matches, PREG_OFFSET_CAPTURE, $tagEndPos + 1)) {
$inner = substr($text, $tagEndPos + 1, $matches[0][1] - $tagEndPos - 1);
$i = $matches[0][1] + strlen($matches[0][0]);
$close = '<close>' . htmlspecialchars($matches[0][0]) . '</close>';
} else {
// No end tag -- let it run out to the end of the text.
$inner = substr($text, $tagEndPos + 1);
$i = strlen($text);
$close = '';
}
}
// <includeonly> and <noinclude> just become <ignore> tags
if (in_array($lowerName, $ignoredElements)) {
$accum .= '<ignore>' . htmlspecialchars(substr($text, $tagStartPos, $i - $tagStartPos)) . '</ignore>';
continue;
}
$accum .= '<ext>';
if ($attrEnd <= $attrStart) {
$attr = '';
} else {
$attr = substr($text, $attrStart, $attrEnd - $attrStart);
}
$accum .= '<name>' . htmlspecialchars($name) . '</name>' . '<attr>' . htmlspecialchars($attr) . '</attr>';
if ($inner !== null) {
$accum .= '<inner>' . htmlspecialchars($inner) . '</inner>';
}
$accum .= $close . '</ext>';
} elseif ($found == 'line-start') {
// Is this the start of a heading?
// Line break belongs before the heading element in any case
if ($fakeLineStart) {
$fakeLineStart = false;
} else {
$accum .= $curChar;
$i++;
}
$count = strspn($text, '=', $i, 6);
if ($count == 1 && $findEquals) {
// DWIM: This looks kind of like a name/value separator
// Let's let the equals handler have it and break the potential heading
// This is heuristic, but AFAICT the methods for completely correct disambiguation are very complex.
} elseif ($count > 0) {
$piece = array('open' => "\n", 'close' => "\n", 'parts' => array(new PPDPart(str_repeat('=', $count))), 'startPos' => $i, 'count' => $count);
$stack->push($piece);
$accum =& $stack->getAccum();
extract($stack->getFlags());
$i += $count;
}
} elseif ($found == 'line-end') {
$piece = $stack->top;
// A heading must be open, otherwise \n wouldn't have been in the search list
assert($piece->open == "\n");
$part = $piece->getCurrentPart();
// Search back through the input to see if it has a proper close
// Do this using the reversed string since the other solutions (end anchor, etc.) are inefficient
$wsLength = strspn($revText, " \t", strlen($text) - $i);
$searchStart = $i - $wsLength;
if (isset($part->commentEnd) && $searchStart - 1 == $part->commentEnd) {
// Comment found at line end
// Search for equals signs before the comment
$searchStart = $part->visualEnd;
$searchStart -= strspn($revText, " \t", strlen($text) - $searchStart);
}
$count = $piece->count;
$equalsLength = strspn($revText, '=', strlen($text) - $searchStart);
if ($equalsLength > 0) {
if ($i - $equalsLength == $piece->startPos) {
// This is just a single string of equals signs on its own line
// Replicate the doHeadings behaviour /={count}(.+)={count}/
// First find out how many equals signs there really are (don't stop at 6)
$count = $equalsLength;
if ($count < 3) {
$count = 0;
} else {
$count = min(6, intval(($count - 1) / 2));
}
} else {
$count = min($equalsLength, $count);
}
if ($count > 0) {
// Normal match, output <h>
$element = "<h level=\"{$count}\" i=\"{$headingIndex}\">{$accum}</h>";
$headingIndex++;
} else {
// Single equals sign on its own line, count=0
$element = $accum;
}
} else {
// No match, no <h>, just pass down the inner text
$element = $accum;
}
// Unwind the stack
$stack->pop();
$accum =& $stack->getAccum();
extract($stack->getFlags());
// Append the result to the enclosing accumulator
$accum .= $element;
// Note that we do NOT increment the input pointer.
// This is because the closing linebreak could be the opening linebreak of
// another heading. Infinite loops are avoided because the next iteration MUST
// hit the heading open case above, which unconditionally increments the
// input pointer.
} elseif ($found == 'open') {
# count opening brace characters
$count = strspn($text, $curChar, $i);
# we need to add to stack only if opening brace count is enough for one of the rules
if ($count >= $rule['min']) {
# Add it to the stack
$piece = array('open' => $curChar, 'close' => $rule['end'], 'count' => $count, 'lineStart' => $i > 0 && $text[$i - 1] == "\n");
$stack->push($piece);
$accum =& $stack->getAccum();
extract($stack->getFlags());
} else {
# Add literal brace(s)
$accum .= htmlspecialchars(str_repeat($curChar, $count));
}
$i += $count;
} elseif ($found == 'close') {
$piece = $stack->top;
# lets check if there are enough characters for closing brace
$maxCount = $piece->count;
$count = strspn($text, $curChar, $i, $maxCount);
# check for maximum matching characters (if there are 5 closing
# characters, we will probably need only 3 - depending on the rules)
$matchingCount = 0;
$rule = $rules[$piece->open];
if ($count > $rule['max']) {
# The specified maximum exists in the callback array, unless the caller
# has made an error
$matchingCount = $rule['max'];
} else {
# Count is less than the maximum
# Skip any gaps in the callback array to find the true largest match
# Need to use array_key_exists not isset because the callback can be null
$matchingCount = $count;
while ($matchingCount > 0 && !array_key_exists($matchingCount, $rule['names'])) {
--$matchingCount;
}
}
if ($matchingCount <= 0) {
# No matching element found in callback array
# Output a literal closing brace and continue
$accum .= htmlspecialchars(str_repeat($curChar, $count));
$i += $count;
continue;
}
$name = $rule['names'][$matchingCount];
if ($name === null) {
// No element, just literal text
$element = $piece->breakSyntax($matchingCount) . str_repeat($rule['end'], $matchingCount);
} else {
# Create XML element
# Note: $parts is already XML, does not need to be encoded further
$parts = $piece->parts;
$title = $parts[0]->out;
unset($parts[0]);
# The invocation is at the start of the line if lineStart is set in
# the stack, and all opening brackets are used up.
if ($maxCount == $matchingCount && !empty($piece->lineStart)) {
$attr = ' lineStart="1"';
} else {
$attr = '';
}
$element = "<{$name}{$attr}>";
$element .= "<title>{$title}</title>";
$argIndex = 1;
foreach ($parts as $partIndex => $part) {
if (isset($part->eqpos)) {
$argName = substr($part->out, 0, $part->eqpos);
$argValue = substr($part->out, $part->eqpos + 1);
$element .= "<part><name>{$argName}</name>=<value>{$argValue}</value></part>";
} else {
$element .= "<part><name index=\"{$argIndex}\" /><value>{$part->out}</value></part>";
$argIndex++;
}
}
$element .= "</{$name}>";
}
# Advance input pointer
$i += $matchingCount;
# Unwind the stack
$stack->pop();
$accum =& $stack->getAccum();
# Re-add the old stack element if it still has unmatched opening characters remaining
if ($matchingCount < $piece->count) {
$piece->parts = array(new PPDPart());
$piece->count -= $matchingCount;
# do we still qualify for any callback with remaining count?
$names = $rules[$piece->open]['names'];
$skippedBraces = 0;
$enclosingAccum =& $accum;
while ($piece->count) {
if (array_key_exists($piece->count, $names)) {
$stack->push($piece);
$accum =& $stack->getAccum();
break;
}
--$piece->count;
$skippedBraces++;
}
$enclosingAccum .= str_repeat($piece->open, $skippedBraces);
}
extract($stack->getFlags());
# Add XML element to the enclosing accumulator
$accum .= $element;
} elseif ($found == 'pipe') {
$findEquals = true;
// shortcut for getFlags()
$stack->addPart();
$accum =& $stack->getAccum();
++$i;
} elseif ($found == 'equals') {
$findEquals = false;
// shortcut for getFlags()
$stack->getCurrentPart()->eqpos = strlen($accum);
$accum .= '=';
++$i;
}
}
# Output any remaining unclosed brackets
foreach ($stack->stack as $piece) {
$stack->rootAccum .= $piece->breakSyntax();
}
$stack->rootAccum .= '</root>';
$xml = $stack->rootAccum;
wfProfileOut(__METHOD__ . '-makexml');
wfProfileIn(__METHOD__ . '-loadXML');
$dom = new DOMDocument();
wfSuppressWarnings();
$result = $dom->loadXML($xml);
wfRestoreWarnings();
if (!$result) {
// Try running the XML through UtfNormal to get rid of invalid characters
$xml = UtfNormal::cleanUp($xml);
$result = $dom->loadXML($xml);
if (!$result) {
throw new MWException(__METHOD__ . ' generated invalid XML');
}
}
$obj = new PPNode_DOM($dom->documentElement);
wfProfileOut(__METHOD__ . '-loadXML');
wfProfileOut(__METHOD__);
return $obj;
}
/**
* @return array
*/
public function getFlags()
{
if (!count($this->stack)) {
return array('findEquals' => false, 'findPipe' => false, 'inHeading' => false);
} else {
return $this->top->getFlags();
}
}
public function __construct()
{
$this->elementClass = 'PPDStackElement_Hash';
parent::__construct();
$this->rootAccum = new PPDAccum_Hash();
}