/**
* @param \Elastica\Document[] $documents
* @param string|null $shardTimeout Timeout in Elasticsearch time format (1m, 15s, ...)
*/
protected function sendDocuments(array $documents, $shardTimeout = null)
{
if (count($documents) === 0) {
return;
}
try {
// addDocuments (notice plural) is the bulk api
$bulk = new \Elastica\Bulk(Connection::getSingleton()->getClient2());
if ($shardTimeout !== null) {
$bulk->setShardTimeout($shardTimeout);
}
$index = Connection::getFlowIndex(wfWikiId());
$type = $index->getType($this->getTypeName());
$bulk->setType($type);
$bulk->addDocuments($documents);
$bulk->send();
} catch (\Exception $e) {
$documentIds = array_map(function ($doc) {
return $doc->getId();
}, $documents);
wfWarn(__METHOD__ . ': Failed updating documents (' . implode(',', $documentIds) . '): ' . $e->getMessage());
}
}
/**
* We want to retrieve the total amount of search word hits
* (static::termsAggregation) but our search terms may not be how
* ElasticSearch stores the words in its index.
* Elastic will "analyze" text (perform stemming, etc) and store
* the terms in a normalized way.
* AFAICT, there is not really a way to get to that information
* from within a search query.
*
* Luckily, since 1.0, Elastic supports _termvector, which gives
* you statistics about the terms in your document.
* Since 1.4, Elastic supports feeding _termvector documents to
* analyze.
* We're going to (ab)use this by letting it respond with term
* information on a bogus document that contains only our current
* search terms.
* So we'll give it a document with just our keywords for the
* column that we're searching in (revisions.text) and Elastic will
* use that column's configuration to analyze the text we feed it.
* It will then respond with the normalized terms & their stats.
*
* @param string $terms
* @return array
*/
protected function getTerms($terms)
{
$terms = preg_split('/\\s+/', $terms);
// _termvectors only works on a type, but our types are
// configured exactly the same so it doesn't matter which
$types = Connection::getAllTypes();
$searchable = Connection::getFlowIndex($this->indexBaseName);
$searchable = $searchable->getType(array_pop($types));
$query = array('doc' => array('revisions' => array('text' => $terms)), "fields" => array("revisions.text"));
// Elastica has no abstraction over _termvector like it has
// for _query, so just do the request ourselves
$response = $searchable->request('_termvector', Request::POST, $query, array());
$data = $response->getData();
return array_keys($data['term_vectors']['revisions.text']['terms']);
}
protected function getOldHeaderType()
{
$index = Connection::getFlowIndex($this->indexBaseName);
return $index->getType(Connection::HEADER_TYPE_NAME);
}