public function execute()
{
global $wgFlowSearchMaintenanceTimeout;
// Set the timeout for maintenance actions
Connection::getSingleton()->setTimeout2($wgFlowSearchMaintenanceTimeout);
/** @var Updater[] $updaters */
$updaters = Container::get('searchindex.updaters');
foreach ($updaters as $updaterType => $updater) {
$fromId = $this->getOption('fromId', null);
$fromId = $fromId ? UUID::create($fromId) : null;
$toId = $this->getOption('toId', null);
$toId = $toId ? UUID::create($toId) : null;
$namespace = $this->getOption('namespace', null);
$numRevisionsToIndex = $this->getOption('limit', null);
$total = 0;
while (true) {
// if a limit was provided, we should make sure to not fetch
// more revisions than asked for
$options = array('LIMIT' => $this->mBatchSize);
if ($numRevisionsToIndex) {
$options['LIMIT'] = min($numRevisionsToIndex, $this->mBatchSize);
// since we do this in batches, we'll subtract the size of
// each batch until $numRevisionsToIndex is reached
$numRevisionsToIndex -= $this->mBatchSize;
if ($options['LIMIT'] <= 0) {
break;
}
}
$conditions = $updater->buildQueryConditions($fromId, $toId, $namespace);
$revisions = $updater->getRevisions($conditions, $options);
// stop if we're all out of revisions
if (!$revisions) {
break;
}
$total += $updater->updateRevisions($revisions, null, null);
$this->output("Indexed {$total} {$updaterType} document(s)\n");
// prepare for next batch, starting at the next id
// prevFromId will default to around unix epoch - there can be
// no data before that
$prevFromId = $fromId ?: UUID::getComparisonUUID('1');
$fromId = $this->getNextFromId($revisions);
// make sure we don't get stuck in an infinite loop
$diff = $prevFromId->getTimestampObj()->diff($fromId->getTimestampObj());
// invert will be 1 if the diff is a negative time period from
// $prevFromId to $fromId, which means that the new $timestamp is
// more recent than our current $result
if ($diff->invert) {
$this->error('Got stuck in an infinite loop.' . "\n" . 'workflow_last_update_timestamp is likely incorrect ' . 'for some workflows.' . "\n" . 'Run maintenance/FlowFixWorkflowLastUpdateTimestamp.php ' . 'to automatically fix those.', 1);
}
// prevent memory from being filled up
Container::get('storage')->clear();
}
}
}
/**
* 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']);
}
/**
* @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());
}
}
/**
* Set the search index to search in.
* false is allowed (means we'll search *all* types)
*
* @param string|false $type
* @throws InvalidInputException
*/
public function setType($type)
{
$allowedTypes = array_merge(Connection::getAllTypes(), array(false));
if (!in_array($type, $allowedTypes)) {
throw new InvalidInputException('Invalid search sort requested', 'invalid-input');
}
$this->type = $type;
}
protected function getAllIndices()
{
return Connection::getAllIndices();
}
public function getParamDescription()
{
$p = $this->getModulePrefix();
return array('term' => 'Search term', 'title' => "Title of the boards to search in. Cannot be used together with {$p}pageid", 'pageid' => "ID of the boards to search in. Cannot be used together with {$p}title", 'namespaces' => 'Namespaces to search in', 'moderationState' => 'Search for revisions in (a) particular moderation state(s)', 'sort' => 'What to order the search results by', 'type' => 'Desired type of results (' . implode('|', Connection::getAllTypes()) . ')', 'offset' => 'Offset value to start fetching results at', 'limit' => 'Amount of results to fetch');
}
