DCausse has uploaded a new change for review.
https://gerrit.wikimedia.org/r/270260
Change subject: Completion suggester: add support for the offset param
......................................................................
Completion suggester: add support for the offset param
The hack is to fetch offset+limit.
Added a hard limit to prevent scrolling to far.
Change-Id: Ia6f5abeb3b623736fcc07df498ddc1cf2691e445
---
M CirrusSearch.php
M includes/CirrusSearch.php
M includes/CompletionSuggester.php
3 files changed, 51 insertions(+), 9 deletions(-)
git pull ssh://gerrit.wikimedia.org:29418/mediawiki/extensions/CirrusSearch
refs/changes/60/270260/1
diff --git a/CirrusSearch.php b/CirrusSearch.php
index e289e78..319ab04 100644
--- a/CirrusSearch.php
+++ b/CirrusSearch.php
@@ -765,6 +765,13 @@
$wgCirrusSearchUseCompletionSuggester = false;
/**
+ * Maximum number of results to ask from the elasticsearch completion
+ * api, note that this value will be multiplied by fetch_limit_factor
+ * set in Completion profiles (default to 2)
+ */
+$wgCirrusSearchCompletionSuggesterHardLimit = 50;
+
+/**
* Try to recycle the completion suggester, if the wiki is small
* it's certainly better to not re-create the index from scratch
* since index creation is costly. Recycling the index will prevent
diff --git a/includes/CirrusSearch.php b/includes/CirrusSearch.php
index ae7d50c..0003250 100644
--- a/includes/CirrusSearch.php
+++ b/includes/CirrusSearch.php
@@ -394,7 +394,8 @@
// offset is omitted, searchSuggestion does not support
// scrolling results
$suggester = new CompletionSuggester( $this->connection,
$this->limit,
- $config, $this->namespaces, null,
$this->indexBaseName );
+ $this->offset, $config, $this->namespaces, null,
+ $this->indexBaseName );
$response = $suggester->suggest( $search, $variants );
if ( $response->isOK() ) {
diff --git a/includes/CompletionSuggester.php b/includes/CompletionSuggester.php
index d256dac..8828da7 100644
--- a/includes/CompletionSuggester.php
+++ b/includes/CompletionSuggester.php
@@ -106,6 +106,11 @@
private $limit;
/**
+ * @var integer offset
+ */
+ private $offset;
+
+ /**
* @var string index base name to use
*/
private $indexBaseName;
@@ -136,7 +141,7 @@
* @param string|boolean $index Base name for index to search from,
defaults to wfWikiId()
* @throws \ConfigException
*/
- public function __construct( Connection $conn, $limit, SearchConfig
$config = null, array $namespaces = null,
+ public function __construct( Connection $conn, $limit, $offset = 0,
SearchConfig $config = null, array $namespaces = null,
User $user = null, $index = false ) {
if ( is_null( $config ) ) {
@@ -148,6 +153,7 @@
parent::__construct( $conn, $user, $config->get(
'CirrusSearchSlowSearch' ) );
$this->config = $config;
$this->limit = $limit;
+ $this->offset = $offset;
$this->indexBaseName = $index ?: $config->getWikiId();
$this->searchContext = new SearchContext( $this->config,
$namespaces );
}
@@ -189,8 +195,7 @@
try {
$result = $index->request( "_suggest",
Request::POST, $suggest, $queryOptions );
if( $result->isOk() ) {
- $result =
$searcher->postProcessSuggest( $result,
- $profiles, $this->limit
);
+ $result =
$searcher->postProcessSuggest( $result, $profiles );
return $searcher->success(
$result );
}
return $result;
@@ -293,7 +298,7 @@
'text' => $query,
'completion' => array(
'field' => $field,
- 'size' => $this->limit *
$config['fetch_limit_factor']
+ 'size' => ($this->limit + $this->offset) *
$config['fetch_limit_factor']
)
);
if ( isset( $config['fuzzy'] ) ) {
@@ -383,11 +388,12 @@
* @param int $limit Maximum suggestions to return, -1 for unlimited
* @return SearchSuggestionSet a set of Suggestions
*/
- protected function postProcessSuggest( \Elastica\Response $response,
$profiles, $limit = -1 ) {
+ protected function postProcessSuggest( \Elastica\Response $response,
$profiles ) {
$this->logContext['elasticTookMs'] = intval(
$response->getQueryTime() * 1000 );
$data = $response->getData();
unset( $data['_shards'] );
+ $limit = $this->getHardLimit();
$suggestions = array();
foreach ( $data as $name => $results ) {
$discount = $profiles[$name]['discount'];
@@ -423,12 +429,14 @@
$this->logContext['hitsTotal'] = count( $suggestions );
- if ( $limit > 0 ) {
- $suggestions = array_slice( $suggestions, 0, $limit,
true );
+ if ( $this->offset < $limit ) {
+ $suggestions = array_slice( $suggestions,
$this->offset, $limit - $this->offset, true );
+ } else {
+ $suggestions = array();
}
$this->logContext['hitsReturned'] = count( $suggestions );
- $this->logContext['hitsOffset'] = 0;
+ $this->logContext['hitsOffset'] = $this->offset;
// we must fetch redirect data for redirect suggestions
$missingText = array();
@@ -501,4 +509,30 @@
public function setLimit( $limit ) {
$this->limit = $limit;
}
+
+ /**
+ * Set the offset
+ * @param int $offset
+ */
+ public function setOffset( $offset ) {
+ $this->offset = $offset;
+ }
+
+ /**
+ * Get the hard limit
+ * The completion api does not supports offset we have to add a hack
+ * here to work around this limitation.
+ * To avoid ridiculously large queris we set also a hard limit.
+ * Note that this limit will be changed by fetch_limit_factor set to 2
or 1.5
+ * depending on the profile.
+ * @return int the number of results to fetch from elastic
+ */
+ private function getHardLimit() {
+ $limit = $this->limit + $this->offset;
+ $hardLimit = $this->config->get(
'CirrusSearchCompletionSuggesterHardLimit' );
+ if ( $limit > $hardLimit ) {
+ return $hardLimit;
+ }
+ return $limit;
+ }
}
--
To view, visit https://gerrit.wikimedia.org/r/270260
To unsubscribe, visit https://gerrit.wikimedia.org/r/settings
Gerrit-MessageType: newchange
Gerrit-Change-Id: Ia6f5abeb3b623736fcc07df498ddc1cf2691e445
Gerrit-PatchSet: 1
Gerrit-Project: mediawiki/extensions/CirrusSearch
Gerrit-Branch: master
Gerrit-Owner: DCausse <dcau...@wikimedia.org>
_______________________________________________
MediaWiki-commits mailing list
MediaWiki-commits@lists.wikimedia.org
https://lists.wikimedia.org/mailman/listinfo/mediawiki-commits
