/**
* Extract parts of the text - opening, main and auxiliary.
*/
private function extractWikitextParts()
{
if (!is_null($this->allText)) {
return;
}
$this->parserOutput->setEditSectionTokens(false);
$this->parserOutput->setTOCEnabled(false);
$text = $this->parserOutput->getText();
if (strlen($text) == 0) {
$this->allText = "";
// empty text - nothing to seek here
return;
}
$opening = null;
$this->openingText = $this->extractHeadingBeforeFirstHeading($text);
// Add extra spacing around break tags so text crammed together like<br>this
// doesn't make one word.
$text = str_replace('<br', "\n<br", $text);
$formatter = new HtmlFormatter($text);
// Strip elements from the page that we never want in the search text.
$formatter->remove($this->excludedElementSelectors);
$formatter->filterContent();
// Strip elements from the page that are auxiliary text. These will still be
// searched but matches will be ranked lower and non-auxiliary matches will be
// preferred in highlighting.
$formatter->remove($this->auxiliaryElementSelectors);
$auxiliaryElements = $formatter->filterContent();
$this->allText = trim(Sanitizer::stripAllTags($formatter->getText()));
foreach ($auxiliaryElements as $auxiliaryElement) {
$this->auxText[] = trim(Sanitizer::stripAllTags($formatter->getText($auxiliaryElement)));
}
}
/**
* This function accomplishes several tasks:
* 1) Auto-number headings if that option is enabled
* 2) Add an [edit] link to sections for users who have enabled the option and can edit the page
* 3) Add a Table of contents on the top for users who have enabled the option
* 4) Auto-anchor headings
*
* It loops through all headlines, collects the necessary data, then splits up the
* string and re-inserts the newly formatted headlines.
*
* @param $text String
* @param string $origText original, untouched wikitext
* @param $isMain Boolean
* @return mixed|string
* @private
*/
function formatHeadings($text, $origText, $isMain = true)
{
global $wgMaxTocLevel, $wgExperimentalHtmlIds;
# Inhibit editsection links if requested in the page
if (isset($this->mDoubleUnderscores['noeditsection'])) {
$maybeShowEditLink = $showEditLink = false;
} else {
$maybeShowEditLink = true;
/* Actual presence will depend on ParserOptions option */
$showEditLink = $this->mOptions->getEditSection();
}
if ($showEditLink) {
$this->mOutput->setEditSectionTokens(true);
}
# Get all headlines for numbering them and adding funky stuff like [edit]
# links - this is for later, but we need the number of headlines right now
$matches = array();
$numMatches = preg_match_all('/<H(?P<level>[1-6])(?P<attrib>.*?' . '>)\\s*(?P<header>[\\s\\S]*?)\\s*<\\/H[1-6] *>/i', $text, $matches);
# if there are fewer than 4 headlines in the article, do not show TOC
# unless it's been explicitly enabled.
$enoughToc = $this->mShowToc && ($numMatches >= 4 || $this->mForceTocPosition);
# Allow user to stipulate that a page should have a "new section"
# link added via __NEWSECTIONLINK__
if (isset($this->mDoubleUnderscores['newsectionlink'])) {
$this->mOutput->setNewSection(true);
}
# Allow user to remove the "new section"
# link via __NONEWSECTIONLINK__
if (isset($this->mDoubleUnderscores['nonewsectionlink'])) {
$this->mOutput->hideNewSection(true);
}
# if the string __FORCETOC__ (not case-sensitive) occurs in the HTML,
# override above conditions and always show TOC above first header
if (isset($this->mDoubleUnderscores['forcetoc'])) {
$this->mShowToc = true;
$enoughToc = true;
}
# headline counter
$headlineCount = 0;
$numVisible = 0;
# Ugh .. the TOC should have neat indentation levels which can be
# passed to the skin functions. These are determined here
$toc = '';
$full = '';
$head = array();
$sublevelCount = array();
$levelCount = array();
$level = 0;
$prevlevel = 0;
$toclevel = 0;
$prevtoclevel = 0;
$markerRegex = "{$this->mUniqPrefix}-h-(\\d+)-" . self::MARKER_SUFFIX;
$baseTitleText = $this->mTitle->getPrefixedDBkey();
$oldType = $this->mOutputType;
$this->setOutputType(self::OT_WIKI);
$frame = $this->getPreprocessor()->newFrame();
$root = $this->preprocessToDom($origText);
$node = $root->getFirstChild();
$byteOffset = 0;
$tocraw = array();
$refers = array();
foreach ($matches[3] as $headline) {
$isTemplate = false;
$titleText = false;
$sectionIndex = false;
$numbering = '';
$markerMatches = array();
if (preg_match("/^{$markerRegex}/", $headline, $markerMatches)) {
$serial = $markerMatches[1];
list($titleText, $sectionIndex) = $this->mHeadings[$serial];
$isTemplate = $titleText != $baseTitleText;
$headline = preg_replace("/^{$markerRegex}\\s*/", "", $headline);
}
if ($toclevel) {
$prevlevel = $level;
}
$level = $matches[1][$headlineCount];
if ($level > $prevlevel) {
# Increase TOC level
$toclevel++;
$sublevelCount[$toclevel] = 0;
if ($toclevel < $wgMaxTocLevel) {
$prevtoclevel = $toclevel;
$toc .= Linker::tocIndent();
$numVisible++;
}
} elseif ($level < $prevlevel && $toclevel > 1) {
# Decrease TOC level, find level to jump to
for ($i = $toclevel; $i > 0; $i--) {
if ($levelCount[$i] == $level) {
# Found last matching level
$toclevel = $i;
break;
} elseif ($levelCount[$i] < $level) {
# Found first matching level below current level
$toclevel = $i + 1;
break;
}
}
if ($i == 0) {
$toclevel = 1;
}
if ($toclevel < $wgMaxTocLevel) {
if ($prevtoclevel < $wgMaxTocLevel) {
# Unindent only if the previous toc level was shown :p
$toc .= Linker::tocUnindent($prevtoclevel - $toclevel);
$prevtoclevel = $toclevel;
} else {
$toc .= Linker::tocLineEnd();
}
}
} else {
# No change in level, end TOC line
if ($toclevel < $wgMaxTocLevel) {
$toc .= Linker::tocLineEnd();
}
}
$levelCount[$toclevel] = $level;
# count number of headlines for each level
$sublevelCount[$toclevel]++;
$dot = 0;
for ($i = 1; $i <= $toclevel; $i++) {
if (!empty($sublevelCount[$i])) {
if ($dot) {
$numbering .= '.';
}
$numbering .= $this->getTargetLanguage()->formatNum($sublevelCount[$i]);
$dot = 1;
}
}
# The safe header is a version of the header text safe to use for links
# Remove link placeholders by the link text.
# <!--LINK number-->
# turns into
# link text with suffix
# Do this before unstrip since link text can contain strip markers
$safeHeadline = $this->replaceLinkHoldersText($headline);
# Avoid insertion of weird stuff like <math> by expanding the relevant sections
$safeHeadline = $this->mStripState->unstripBoth($safeHeadline);
# Strip out HTML (first regex removes any tag not allowed)
# Allowed tags are:
# * <sup> and <sub> (bug 8393)
# * <i> (bug 26375)
# * <b> (r105284)
# * <span dir="rtl"> and <span dir="ltr"> (bug 35167)
#
# We strip any parameter from accepted tags (second regex), except dir="rtl|ltr" from <span>,
# to allow setting directionality in toc items.
$tocline = preg_replace(array('#<(?!/?(span|sup|sub|i|b)(?: [^>]*)?>).*?' . '>#', '#<(/?(?:span(?: dir="(?:rtl|ltr)")?|sup|sub|i|b))(?: .*?)?' . '>#'), array('', '<$1>'), $safeHeadline);
$tocline = trim($tocline);
# For the anchor, strip out HTML-y stuff period
$safeHeadline = preg_replace('/<.*?' . '>/', '', $safeHeadline);
$safeHeadline = Sanitizer::normalizeSectionNameWhitespace($safeHeadline);
# Save headline for section edit hint before it's escaped
$headlineHint = $safeHeadline;
if ($wgExperimentalHtmlIds) {
# For reverse compatibility, provide an id that's
# HTML4-compatible, like we used to.
#
# It may be worth noting, academically, that it's possible for
# the legacy anchor to conflict with a non-legacy headline
# anchor on the page. In this case likely the "correct" thing
# would be to either drop the legacy anchors or make sure
# they're numbered first. However, this would require people
# to type in section names like "abc_.D7.93.D7.90.D7.A4"
# manually, so let's not bother worrying about it.
$legacyHeadline = Sanitizer::escapeId($safeHeadline, array('noninitial', 'legacy'));
$safeHeadline = Sanitizer::escapeId($safeHeadline);
if ($legacyHeadline == $safeHeadline) {
# No reason to have both (in fact, we can't)
$legacyHeadline = false;
}
} else {
$legacyHeadline = false;
$safeHeadline = Sanitizer::escapeId($safeHeadline, 'noninitial');
}
# HTML names must be case-insensitively unique (bug 10721).
# This does not apply to Unicode characters per
# http://dev.w3.org/html5/spec/infrastructure.html#case-sensitivity-and-string-comparison
# @todo FIXME: We may be changing them depending on the current locale.
$arrayKey = strtolower($safeHeadline);
if ($legacyHeadline === false) {
$legacyArrayKey = false;
} else {
$legacyArrayKey = strtolower($legacyHeadline);
}
# count how many in assoc. array so we can track dupes in anchors
if (isset($refers[$arrayKey])) {
$refers[$arrayKey]++;
} else {
$refers[$arrayKey] = 1;
}
if (isset($refers[$legacyArrayKey])) {
$refers[$legacyArrayKey]++;
} else {
$refers[$legacyArrayKey] = 1;
}
# Don't number the heading if it is the only one (looks silly)
if (count($matches[3]) > 1 && $this->mOptions->getNumberHeadings()) {
# the two are different if the line contains a link
$headline = Html::element('span', array('class' => 'mw-headline-number'), $numbering) . ' ' . $headline;
}
# Create the anchor for linking from the TOC to the section
$anchor = $safeHeadline;
$legacyAnchor = $legacyHeadline;
if ($refers[$arrayKey] > 1) {
$anchor .= '_' . $refers[$arrayKey];
}
if ($legacyHeadline !== false && $refers[$legacyArrayKey] > 1) {
$legacyAnchor .= '_' . $refers[$legacyArrayKey];
}
if ($enoughToc && (!isset($wgMaxTocLevel) || $toclevel < $wgMaxTocLevel)) {
$toc .= Linker::tocLine($anchor, $tocline, $numbering, $toclevel, $isTemplate ? false : $sectionIndex);
}
# Add the section to the section tree
# Find the DOM node for this header
$noOffset = $isTemplate || $sectionIndex === false;
while ($node && !$noOffset) {
if ($node->getName() === 'h') {
$bits = $node->splitHeading();
if ($bits['i'] == $sectionIndex) {
break;
}
}
$byteOffset += mb_strlen($this->mStripState->unstripBoth($frame->expand($node, PPFrame::RECOVER_ORIG)));
$node = $node->getNextSibling();
}
$tocraw[] = array('toclevel' => $toclevel, 'level' => $level, 'line' => $tocline, 'number' => $numbering, 'index' => ($isTemplate ? 'T-' : '') . $sectionIndex, 'fromtitle' => $titleText, 'byteoffset' => $noOffset ? null : $byteOffset, 'anchor' => $anchor);
# give headline the correct <h#> tag
if ($maybeShowEditLink && $sectionIndex !== false) {
// Output edit section links as markers with styles that can be customized by skins
if ($isTemplate) {
# Put a T flag in the section identifier, to indicate to extractSections()
# that sections inside <includeonly> should be counted.
$editlinkArgs = array($titleText, "T-{$sectionIndex}");
} else {
$editlinkArgs = array($this->mTitle->getPrefixedText(), $sectionIndex, $headlineHint);
}
// We use a bit of pesudo-xml for editsection markers. The language converter is run later on
// Using a UNIQ style marker leads to the converter screwing up the tokens when it converts stuff
// And trying to insert strip tags fails too. At this point all real inputted tags have already been escaped
// so we don't have to worry about a user trying to input one of these markers directly.
// We use a page and section attribute to stop the language converter from converting these important bits
// of data, but put the headline hint inside a content block because the language converter is supposed to
// be able to convert that piece of data.
$editlink = '<mw:editsection page="' . htmlspecialchars($editlinkArgs[0]);
$editlink .= '" section="' . htmlspecialchars($editlinkArgs[1]) . '"';
if (isset($editlinkArgs[2])) {
$editlink .= '>' . $editlinkArgs[2] . '</mw:editsection>';
} else {
$editlink .= '/>';
}
} else {
$editlink = '';
}
$head[$headlineCount] = Linker::makeHeadline($level, $matches['attrib'][$headlineCount], $anchor, $headline, $editlink, $legacyAnchor);
$headlineCount++;
}
$this->setOutputType($oldType);
# Never ever show TOC if no headers
if ($numVisible < 1) {
$enoughToc = false;
}
if ($enoughToc) {
if ($prevtoclevel > 0 && $prevtoclevel < $wgMaxTocLevel) {
$toc .= Linker::tocUnindent($prevtoclevel - 1);
}
$toc = Linker::tocList($toc, $this->mOptions->getUserLangObj());
$this->mOutput->setTOCHTML($toc);
$toc = self::TOC_START . $toc . self::TOC_END;
}
if ($isMain) {
$this->mOutput->setSections($tocraw);
}
# split up and insert constructed headlines
$blocks = preg_split('/<H[1-6].*?' . '>[\\s\\S]*?<\\/H[1-6]>/i', $text);
$i = 0;
// build an array of document sections
$sections = array();
foreach ($blocks as $block) {
// $head is zero-based, sections aren't.
if (empty($head[$i - 1])) {
$sections[$i] = $block;
} else {
$sections[$i] = $head[$i - 1] . $block;
}
/**
* Send a hook, one per section.
* The idea here is to be able to make section-level DIVs, but to do so in a
* lower-impact, more correct way than r50769
*
* $this : caller
* $section : the section number
* &$sectionContent : ref to the content of the section
* $showEditLinks : boolean describing whether this section has an edit link
*/
wfRunHooks('ParserSectionCreate', array($this, $i, &$sections[$i], $showEditLink));
$i++;
}
if ($enoughToc && $isMain && !$this->mForceTocPosition) {
// append the TOC at the beginning
// Top anchor now in skin
$sections[0] = $sections[0] . $toc . "\n";
}
$full .= join('', $sections);
if ($this->mForceTocPosition) {
return str_replace('<!--MWTOC-->', $toc, $full);
} else {
return $full;
}
}
