Merge "Don't check namespace in SpecialWantedtemplates"
[lhc/web/wiklou.git] / includes / jobqueue / jobs / RecentChangesUpdateJob.php
index b550f50..d6fa26b 100644 (file)
@@ -27,7 +27,7 @@
  * @since 1.25
  */
 class RecentChangesUpdateJob extends Job {
-       function __construct( $title, $params ) {
+       function __construct( Title $title, array $params ) {
                parent::__construct( 'recentChangesUpdate', $title, $params );
 
                if ( !isset( $params['type'] ) ) {
@@ -46,11 +46,24 @@ class RecentChangesUpdateJob extends Job {
                );
        }
 
+       /**
+        * @return RecentChangesUpdateJob
+        * @since 1.26
+        */
+       final public static function newCacheUpdateJob() {
+               return new self(
+                       SpecialPage::getTitleFor( 'Recentchanges' ), array( 'type' => 'cacheUpdate' )
+               );
+       }
+
        public function run() {
                if ( $this->params['type'] === 'purge' ) {
                        $this->purgeExpiredRows();
+               } elseif ( $this->params['type'] === 'cacheUpdate' ) {
+                       $this->updateActiveUsers();
                } else {
-                       throw new Exception( "Invalid 'type' parameter '{$this->params['type']}'." );
+                       throw new InvalidArgumentException(
+                               "Invalid 'type' parameter '{$this->params['type']}'." );
                }
 
                return true;
@@ -62,11 +75,13 @@ class RecentChangesUpdateJob extends Job {
                $lockKey = wfWikiID() . ':recentchanges-prune';
 
                $dbw = wfGetDB( DB_MASTER );
-               if ( !$dbw->lock( $lockKey, __METHOD__, 1 ) ) {
+               if ( !$dbw->lockIsFree( $lockKey, __METHOD__ )
+                       || !$dbw->lock( $lockKey, __METHOD__, 1 )
+               ) {
                        return; // already in progress
                }
-               $batchSize = 100; // Avoid slave lag
 
+               $batchSize = 100; // avoid slave lag
                $cutoff = $dbw->timestamp( time() - $wgRCMaxAge );
                do {
                        $rcIds = $dbw->selectFieldValues( 'recentchanges',
@@ -78,7 +93,7 @@ class RecentChangesUpdateJob extends Job {
                        if ( $rcIds ) {
                                $dbw->delete( 'recentchanges', array( 'rc_id' => $rcIds ), __METHOD__ );
                        }
-                       // No need for this to be in a transaction.
+                       // Commit in chunks to avoid slave lag
                        $dbw->commit( __METHOD__, 'flush' );
 
                        if ( count( $rcIds ) === $batchSize ) {
@@ -92,4 +107,119 @@ class RecentChangesUpdateJob extends Job {
 
                $dbw->unlock( $lockKey, __METHOD__ );
        }
+
+       protected function updateActiveUsers() {
+               global $wgActiveUserDays;
+
+               // Users that made edits at least this many days ago are "active"
+               $days = $wgActiveUserDays;
+               // Pull in the full window of active users in this update
+               $window = $wgActiveUserDays * 86400;
+
+               $dbw = wfGetDB( DB_MASTER );
+               // JobRunner uses DBO_TRX, but doesn't call begin/commit itself;
+               // onTransactionIdle() will run immediately since there is no trx.
+               $dbw->onTransactionIdle( function() use ( $dbw, $days, $window ) {
+                       // Avoid disconnect/ping() cycle that makes locks fall off
+                       $dbw->setSessionOptions( array( 'connTimeout' => 900 ) );
+
+                       $lockKey = wfWikiID() . '-activeusers';
+                       if ( !$dbw->lock( $lockKey, __METHOD__, 1 ) ) {
+                               return false; // exclusive update (avoids duplicate entries)
+                       }
+
+                       $nowUnix = time();
+                       // Get the last-updated timestamp for the cache
+                       $cTime = $dbw->selectField( 'querycache_info',
+                               'qci_timestamp',
+                               array( 'qci_type' => 'activeusers' )
+                       );
+                       $cTimeUnix = $cTime ? wfTimestamp( TS_UNIX, $cTime ) : 1;
+
+                       // Pick the date range to fetch from. This is normally from the last
+                       // update to till the present time, but has a limited window for sanity.
+                       // If the window is limited, multiple runs are need to fully populate it.
+                       $sTimestamp = max( $cTimeUnix, $nowUnix - $days * 86400 );
+                       $eTimestamp = min( $sTimestamp + $window, $nowUnix );
+
+                       // Get all the users active since the last update
+                       $res = $dbw->select(
+                               array( 'recentchanges' ),
+                               array( 'rc_user_text', 'lastedittime' => 'MAX(rc_timestamp)' ),
+                               array(
+                                       'rc_user > 0', // actual accounts
+                                       'rc_type != ' . $dbw->addQuotes( RC_EXTERNAL ), // no wikidata
+                                       'rc_log_type IS NULL OR rc_log_type != ' . $dbw->addQuotes( 'newusers' ),
+                                       'rc_timestamp >= ' . $dbw->addQuotes( $dbw->timestamp( $sTimestamp ) ),
+                                       'rc_timestamp <= ' . $dbw->addQuotes( $dbw->timestamp( $eTimestamp ) )
+                               ),
+                               __METHOD__,
+                               array(
+                                       'GROUP BY' => array( 'rc_user_text' ),
+                                       'ORDER BY' => 'NULL' // avoid filesort
+                               )
+                       );
+                       $names = array();
+                       foreach ( $res as $row ) {
+                               $names[$row->rc_user_text] = $row->lastedittime;
+                       }
+
+                       // Rotate out users that have not edited in too long (according to old data set)
+                       $dbw->delete( 'querycachetwo',
+                               array(
+                                       'qcc_type' => 'activeusers',
+                                       'qcc_value < ' . $dbw->addQuotes( $nowUnix - $days * 86400 ) // TS_UNIX
+                               ),
+                               __METHOD__
+                       );
+
+                       // Find which of the recently active users are already accounted for
+                       if ( count( $names ) ) {
+                               $res = $dbw->select( 'querycachetwo',
+                                       array( 'user_name' => 'qcc_title' ),
+                                       array(
+                                               'qcc_type' => 'activeusers',
+                                               'qcc_namespace' => NS_USER,
+                                               'qcc_title' => array_keys( $names ) ),
+                                       __METHOD__
+                               );
+                               foreach ( $res as $row ) {
+                                       unset( $names[$row->user_name] );
+                               }
+                       }
+
+                       // Insert the users that need to be added to the list
+                       if ( count( $names ) ) {
+                               $newRows = array();
+                               foreach ( $names as $name => $lastEditTime ) {
+                                       $newRows[] = array(
+                                               'qcc_type' => 'activeusers',
+                                               'qcc_namespace' => NS_USER,
+                                               'qcc_title' => $name,
+                                               'qcc_value' => wfTimestamp( TS_UNIX, $lastEditTime ),
+                                               'qcc_namespacetwo' => 0, // unused
+                                               'qcc_titletwo' => '' // unused
+                                       );
+                               }
+                               foreach ( array_chunk( $newRows, 500 ) as $rowBatch ) {
+                                       $dbw->insert( 'querycachetwo', $rowBatch, __METHOD__ );
+                                       wfWaitForSlaves();
+                               }
+                       }
+
+                       // If a transaction was already started, it might have an old
+                       // snapshot, so kludge the timestamp range back as needed.
+                       $asOfTimestamp = min( $eTimestamp, (int)$dbw->trxTimestamp() );
+
+                       // Touch the data freshness timestamp
+                       $dbw->replace( 'querycache_info',
+                               array( 'qci_type' ),
+                               array( 'qci_type' => 'activeusers',
+                                       'qci_timestamp' => $dbw->timestamp( $asOfTimestamp ) ), // not always $now
+                               __METHOD__
+                       );
+
+                       $dbw->unlock( $lockKey, __METHOD__ );
+               } );
+       }
 }