From c1f9b1f7b1b77776192048005dcc66dcf3df2bfb Mon Sep 17 00:00:00 2001 From: Pierre Schmitz Date: Sat, 27 Dec 2014 15:41:37 +0100 Subject: Update to MediaWiki 1.24.1 --- includes/jobqueue/jobs/RefreshLinksJob2.php | 141 ++++++++++++++++++++++++++++ 1 file changed, 141 insertions(+) create mode 100644 includes/jobqueue/jobs/RefreshLinksJob2.php (limited to 'includes/jobqueue/jobs/RefreshLinksJob2.php') diff --git a/includes/jobqueue/jobs/RefreshLinksJob2.php b/includes/jobqueue/jobs/RefreshLinksJob2.php new file mode 100644 index 00000000..97405aeb --- /dev/null +++ b/includes/jobqueue/jobs/RefreshLinksJob2.php @@ -0,0 +1,141 @@ +removeDuplicates = !isset( $params['start'] ) && !isset( $params['end'] ); + } + + /** + * Run a refreshLinks2 job + * @return bool Success + */ + function run() { + global $wgUpdateRowsPerJob; + + $linkCache = LinkCache::singleton(); + $linkCache->clear(); + + if ( is_null( $this->title ) ) { + $this->error = "refreshLinks2: Invalid title"; + return false; + } + + // Back compat for pre-r94435 jobs + $table = isset( $this->params['table'] ) ? $this->params['table'] : 'templatelinks'; + + // Avoid slave lag when fetching templates. + // When the outermost job is run, we know that the caller that enqueued it must have + // committed the relevant changes to the DB by now. At that point, record the master + // position and pass it along as the job recursively breaks into smaller range jobs. + // Hopefully, when leaf jobs are popped, the slaves will have reached that position. + if ( isset( $this->params['masterPos'] ) ) { + $masterPos = $this->params['masterPos']; + } elseif ( wfGetLB()->getServerCount() > 1 ) { + $masterPos = wfGetLB()->getMasterPos(); + } else { + $masterPos = false; + } + + $tbc = $this->title->getBacklinkCache(); + + $jobs = array(); // jobs to insert + if ( isset( $this->params['start'] ) && isset( $this->params['end'] ) ) { + # This is a partition job to trigger the insertion of leaf jobs... + $jobs = array_merge( $jobs, $this->getSingleTitleJobs( $table, $masterPos ) ); + } else { + # This is a base job to trigger the insertion of partitioned jobs... + if ( $tbc->getNumLinks( $table, $wgUpdateRowsPerJob + 1 ) <= $wgUpdateRowsPerJob ) { + # Just directly insert the single per-title jobs + $jobs = array_merge( $jobs, $this->getSingleTitleJobs( $table, $masterPos ) ); + } else { + # Insert the partition jobs to make per-title jobs + foreach ( $tbc->partition( $table, $wgUpdateRowsPerJob ) as $batch ) { + list( $start, $end ) = $batch; + $jobs[] = new RefreshLinksJob2( $this->title, + array( + 'table' => $table, + 'start' => $start, + 'end' => $end, + 'masterPos' => $masterPos, + ) + $this->getRootJobParams() // carry over information for de-duplication + ); + } + } + } + + if ( count( $jobs ) ) { + JobQueueGroup::singleton()->push( $jobs ); + } + + return true; + } + + /** + * @param string $table + * @param mixed $masterPos + * @return array + */ + protected function getSingleTitleJobs( $table, $masterPos ) { + # The "start"/"end" fields are not set for the base jobs + $start = isset( $this->params['start'] ) ? $this->params['start'] : false; + $end = isset( $this->params['end'] ) ? $this->params['end'] : false; + $titles = $this->title->getBacklinkCache()->getLinks( $table, $start, $end ); + # Convert into single page refresh links jobs. + # This handles well when in sapi mode and is useful in any case for job + # de-duplication. If many pages use template A, and that template itself + # uses template B, then an edit to both will create many duplicate jobs. + # Roughly speaking, for each page, one of the "RefreshLinksJob" jobs will + # get run first, and when it does, it will remove the duplicates. Of course, + # one page could have its job popped when the other page's job is still + # buried within the logic of a refreshLinks2 job. + $jobs = array(); + foreach ( $titles as $title ) { + $jobs[] = new RefreshLinksJob( $title, + array( 'masterPos' => $masterPos ) + $this->getRootJobParams() + ); // carry over information for de-duplication + } + return $jobs; + } + + /** + * @return array + */ + public function getDeduplicationInfo() { + $info = parent::getDeduplicationInfo(); + // Don't let highly unique "masterPos" values ruin duplicate detection + if ( is_array( $info['params'] ) ) { + unset( $info['params']['masterPos'] ); + } + return $info; + } +} -- cgit v1.2.3-54-g00ecf