/**
* Build the result.
* @param $results \Elastica\ResultSet containing all search results
* @param $result \Elastica\Result containing the given search result
* @param string $interwiki Interwiki prefix, if any
* @param $result \Elastic\Result containing information about the result this class should represent
*/
public function __construct($results, $result, $interwiki = '')
{
if ($interwiki) {
$this->setInterwiki($result, $interwiki);
}
$this->docId = $result->getId();
$this->mTitle = Title::makeTitle($result->namespace, $result->title, '', $this->interwiki);
if ($this->getTitle()->getNamespace() == NS_FILE) {
$this->mImage = wfFindFile($this->mTitle);
}
$fields = $result->getFields();
// Not all results requested a word count. Just pretend we have none if so
$this->wordCount = isset($fields['text.word_count']) ? $fields['text.word_count'][0] : 0;
$this->byteSize = $result->text_bytes;
$this->timestamp = new MWTimestamp($result->timestamp);
$highlights = $result->getHighlights();
if (isset($highlights['title'])) {
$nstext = $this->getTitle()->getNamespace() === 0 ? '' : Util::getNamespaceText($this->getTitle()) . ':';
$this->titleSnippet = $nstext . $this->escapeHighlightedText($highlights['title'][0]);
} elseif ($this->mTitle->isExternal()) {
// Interwiki searches are weird. They won't have title highlights by design, but
// if we don't return a title snippet we'll get weird display results.
$nsText = $this->getInterwikiNamespaceText();
$titleText = $this->mTitle->getText();
$this->titleSnippet = $nsText ? "{$nsText}:{$titleText}" : $titleText;
}
if (!isset($highlights['title']) && isset($highlights['redirect.title'])) {
// Make sure to find the redirect title before escaping because escaping breaks it....
$redirects = $result->redirect;
$this->redirectTitle = $this->findRedirectTitle($highlights['redirect.title'][0], $redirects);
$this->redirectSnipppet = $this->escapeHighlightedText($highlights['redirect.title'][0]);
}
$this->textSnippet = $this->escapeHighlightedText($this->pickTextSnippet($highlights));
if (isset($highlights['heading'])) {
$this->sectionSnippet = $this->escapeHighlightedText($highlights['heading'][0]);
$this->sectionTitle = $this->findSectionTitle();
}
if (isset($highlights['category'])) {
$this->categorySnippet = $this->escapeHighlightedText($highlights['category'][0]);
}
}
/**
* @param \WikiPage[] $pages
* @param int $flags
*/
private function buildDocumentsForPages($pages, $flags)
{
global $wgCirrusSearchUpdateConflictRetryCount;
$indexOnSkip = $flags & self::INDEX_ON_SKIP;
$skipParse = $flags & self::SKIP_PARSE;
$skipLinks = $flags & self::SKIP_LINKS;
$forceParse = $flags & self::FORCE_PARSE;
$fullDocument = !($skipParse || $skipLinks);
$documents = array();
foreach ($pages as $page) {
$title = $page->getTitle();
if (!$page->exists()) {
LoggerFactory::getInstance('CirrusSearch')->warning('Attempted to build a document for a page that doesn\'t exist. This should be caught ' . "earlier but wasn't. Page: {title}", array('title' => $title));
continue;
}
$doc = new \Elastica\Document($page->getId(), array('version' => $page->getLatest(), 'version_type' => 'external', 'namespace' => $title->getNamespace(), 'namespace_text' => Util::getNamespaceText($title), 'title' => $title->getText(), 'timestamp' => wfTimestamp(TS_ISO_8601, $page->getTimestamp())));
// Everything as sent as an update to prevent overwriting fields maintained in other processes like
// OtherIndex::updateOtherIndex.
// But we need a way to index documents that don't already exist. We're willing to upsert any full
// documents or any documents that we've been explicitly told it is ok to index when they aren't full.
// This is typically just done during the first phase of the initial index build.
// A quick note about docAsUpsert's merging behavior: It overwrites all fields provided by doc unless they
// are objects in both doc and the indexed source. We're ok with this because all of our fields are either
// regular types or lists of objects and lists are overwritten.
$doc->setDocAsUpsert($fullDocument || $indexOnSkip);
$doc->setRetryOnConflict($wgCirrusSearchUpdateConflictRetryCount);
if (!$skipParse) {
// Get text to index, based on content and parser output
list($content, $parserOutput) = $this->getContentAndParserOutput($page, $forceParse);
// Build our page data
$pageBuilder = new PageDataBuilder($doc, $title, $content, $parserOutput);
$doc = $pageBuilder->build();
// And build the page text itself
$textBuilder = new PageTextBuilder($doc, $content, $parserOutput);
$doc = $textBuilder->build();
// If we're a file, build its metadata too
if ($title->getNamespace() === NS_FILE) {
$fileBuilder = new FileDataBuilder($doc, $title);
$doc = $fileBuilder->build();
}
// Then let hooks have a go
MWHooks::run('CirrusSearchBuildDocumentParse', array($doc, $title, $content, $parserOutput, $this->connection));
}
if (!$skipLinks) {
MWHooks::run('CirrusSearchBuildDocumentLinks', array($doc, $title, $this->connection));
}
$documents[] = $doc;
}
MWHooks::run('CirrusSearchBuildDocumentFinishBatch', array($pages));
return $documents;
}
