refreshLinks.php

Go to the documentation of this file.

<?php
use MediaWiki\Linker\LinkTarget;
use MediaWiki\MediaWikiServices;
use MediaWiki\Revision\RevisionRecord;
use MediaWiki\Title\Title;
use Wikimedia\Rdbms\IReadableDatabase;
use Wikimedia\Rdbms\SelectQueryBuilder;
 
require_once __DIR__ . '/Maintenance.php';
 
class RefreshLinks extends Maintenance {
    private const REPORTING_INTERVAL = 100;
 
    public function __construct() {
        parent::__construct();
        $this->addDescription( 'Refresh link tables' );
        $this->addOption( 'verbose', 'Output information about link refresh progress', false, false, 'v' );
        $this->addOption( 'dfn-only', 'Delete links from nonexistent articles only' );
        $this->addOption( 'new-only', 'Only affect articles with just a single edit' );
        $this->addOption( 'redirects-only', 'Only fix redirects, not all links' );
        $this->addOption( 'old-redirects-only', 'Only fix redirects with no redirect table entry' );
        $this->addOption( 'touched-only', 'Only fix pages that have been touched after last update' );
        $this->addOption( 'e', 'Last page id to refresh', false, true );
        $this->addOption( 'dfn-chunk-size', 'Maximum number of existent IDs to check per ' .
            'query, default 100000', false, true );
        $this->addOption( 'namespace', 'Only fix pages in this namespace', false, true );
        $this->addOption( 'category', 'Only fix pages in this category', false, true );
        $this->addOption( 'tracking-category', 'Only fix pages in this tracking category', false, true );
        $this->addOption( 'before-timestamp', 'Only fix pages that were last updated before this timestamp',
            false, true );
        $this->addArg( 'start', 'Page_id to start from, default 1', false );
        $this->setBatchSize( 100 );
    }
 
    public function execute() {
        // Note that there is a difference between not specifying the start
        // and end IDs and using the minimum and maximum values from the page
        // table. In the latter case, deleteLinksFromNonexistent() will not
        // delete entries for nonexistent IDs that fall outside the range.
        $start = (int)$this->getArg( 0 ) ?: null;
        $end = (int)$this->getOption( 'e' ) ?: null;
        $dfnChunkSize = (int)$this->getOption( 'dfn-chunk-size', 100000 );
 
        if ( $this->hasOption( 'dfn-only' ) ) {
            $this->deleteLinksFromNonexistent( $start, $end, $this->getBatchSize(), $dfnChunkSize );
            return;
        }
 
        $dbr = $this->getDB( DB_REPLICA, [ 'vslow' ] );
        $builder = $dbr->newSelectQueryBuilder()
            ->from( 'page' )
            ->where( self::intervalCond( $dbr, 'page_id', $start, $end ) )
            ->limit( $this->getBatchSize() );
 
        if ( $this->hasOption( 'namespace' ) ) {
            $builder->andWhere( [ 'page_namespace' => (int)$this->getOption( 'namespace' ) ] );
        }
 
        if ( $this->hasOption( 'before-timestamp' ) ) {
            $timeCond = $dbr->buildComparison( '<', [
                'page_links_updated' => $this->getOption( 'before-timestamp' )
            ] );
            $builder->andWhere( [ "$timeCond OR page_links_updated IS NULL" ] );
        }
 
        if ( $this->hasOption( 'category' ) ) {
            $category = $this->getOption( 'category' );
            $title = Title::makeTitleSafe( NS_CATEGORY, $category );
            if ( !$title ) {
                $this->fatalError( "'$category' is an invalid category name!\n" );
            }
            $this->refreshCategory( $builder, $title );
        } elseif ( $this->hasOption( 'tracking-category' ) ) {
            // See TrackingCategories::CORE_TRACKING_CATEGORIES for tracking category keys defined by core
            $this->refreshTrackingCategory( $builder, $this->getOption( 'tracking-category' ) );
        } else {
            $new = $this->hasOption( 'new-only' );
            $redir = $this->hasOption( 'redirects-only' );
            $oldRedir = $this->hasOption( 'old-redirects-only' );
            $touched = $this->hasOption( 'touched-only' );
            $what = $redir ? 'redirects' : 'links';
            if ( $oldRedir ) {
                $builder->leftJoin( 'redirect', null, 'page_id=rd_from' )
                    ->andWhere( [
                        'page_is_redirect' => 1,
                        'rd_from' => null,
                    ] );
                $this->output( "Refreshing old redirects from $start...\n" );
            } elseif ( $new ) {
                $builder->andWhere( [ 'page_is_new' => 1 ] );
                $this->output( "Refreshing $what from new pages...\n" );
            } else {
                if ( $touched ) {
                    $builder->andWhere( [
                        'page_touched > page_links_updated OR page_links_updated IS NULL',
                    ] );
                }
                $this->output( "Refreshing $what from pages...\n" );
            }
            $this->doRefreshLinks( $builder, $redir || $oldRedir );
            if ( !$this->hasOption( 'namespace' ) ) {
                $this->deleteLinksFromNonexistent( $start, $end, $this->getBatchSize(), $dfnChunkSize );
            }
        }
    }
 
    private function doRefreshLinks(
        SelectQueryBuilder $builder,
        bool $redirectsOnly = false,
        array $indexFields = [ 'page_id' ]
    ) {
        // Give extensions a chance to optimize settings
        $this->getHookRunner()->onMaintenanceRefreshLinksInit( $this );
 
        $estimateCount = $builder->estimateRowCount();
        $this->output( "Estimated page count: $estimateCount\n" );
 
        $i = 0;
        $lastIndexes = array_fill_keys( $indexFields, 0 );
        $selectFields = in_array( 'page_id', $indexFields )
            ? $indexFields : [ 'page_id', ...$indexFields ];
        $verbose = $this->hasOption( 'verbose' );
        $dbr = $this->getDB( DB_REPLICA, [ 'vslow' ] );
        do {
            $batchCond = $dbr->buildComparison( '>', $lastIndexes );
            $res = ( clone $builder )->select( $selectFields )
                ->andWhere( [ $batchCond ] )
                ->orderBy( $indexFields )
                ->caller( __METHOD__ )->fetchResultSet();
 
            if ( $verbose ) {
                $this->output( "Refreshing links for {$res->numRows()} pages\n" );
            }
 
            foreach ( $res as $row ) {
                if ( !( ++$i % self::REPORTING_INTERVAL ) ) {
                    $this->output( "$i\n" );
                    $this->waitForReplication();
                }
                if ( $verbose ) {
                    $this->output( "Refreshing links for page ID {$row->page_id}\n" );
                }
                self::fixRedirect( $this, $row->page_id );
                if ( !$redirectsOnly ) {
                    self::fixLinksFromArticle( $row->page_id );
                }
            }
            if ( $res->numRows() ) {
                $res->seek( $res->numRows() - 1 );
                foreach ( $indexFields as $field ) {
                    $lastIndexes[$field] = $res->current()->$field;
                }
            }
 
        } while ( $res->numRows() == $this->getBatchSize() );
    }
 
    public static function fixRedirect( Maintenance $maint, $id ) {
        $page = $maint->getServiceContainer()->getWikiPageFactory()->newFromID( $id );
 
        // In case the page just got deleted.
        if ( $page === null ) {
            return;
        }
 
        $rt = null;
        $content = $page->getContent( RevisionRecord::RAW );
        if ( $content !== null ) {
            $rt = $content->getRedirectTarget();
        }
 
        $dbw = $maint->getDB( DB_PRIMARY );
        if ( $rt === null ) {
            // The page is not a redirect
            // Delete any redirect table entry for it
            $dbw->delete( 'redirect', [ 'rd_from' => $id ], __METHOD__ );
            $fieldValue = 0;
        } else {
            $page->insertRedirectEntry( $rt );
            $fieldValue = 1;
        }
 
        // Update the page table to be sure it is an a consistent state
        $dbw->update( 'page', [ 'page_is_redirect' => $fieldValue ],
            [ 'page_id' => $id ], __METHOD__ );
    }
 
    public static function fixLinksFromArticle( $id ) {
        $services = MediaWikiServices::getInstance();
        $page = $services->getWikiPageFactory()->newFromID( $id );
 
        // In case the page just got deleted.
        if ( $page === null ) {
            return;
        }
 
        // Defer updates to post-send but then immediately execute deferred updates;
        // this is the simplest way to run all updates immediately (including updates
        // scheduled by other updates).
        $page->doSecondaryDataUpdates( [
            'defer' => DeferredUpdates::POSTSEND,
            'causeAction' => 'refresh-links-maintenance',
            'recursive' => false,
        ] );
        DeferredUpdates::doUpdates();
    }
 
    private function deleteLinksFromNonexistent( $start = null, $end = null, $batchSize = 100,
        $chunkSize = 100000
    ) {
        $this->waitForReplication();
        $this->output( "Deleting illegal entries from the links tables...\n" );
        $dbr = $this->getDB( DB_REPLICA, [ 'vslow' ] );
        do {
            // Find the start of the next chunk. This is based only
            // on existent page_ids.
            $nextStart = $dbr->newSelectQueryBuilder()
                ->select( 'page_id' )
                ->from( 'page' )
                ->where( [ self::intervalCond( $dbr, 'page_id', $start, $end ) ] )
                ->orderBy( 'page_id' )
                ->offset( $chunkSize )
                ->caller( __METHOD__ )->fetchField();
 
            if ( $nextStart !== false ) {
                // To find the end of the current chunk, subtract one.
                // This will serve to limit the number of rows scanned in
                // dfnCheckInterval(), per query, to at most the sum of
                // the chunk size and deletion batch size.
                $chunkEnd = $nextStart - 1;
            } else {
                // This is the last chunk. Check all page_ids up to $end.
                $chunkEnd = $end;
            }
 
            $fmtStart = $start !== null ? "[$start" : '(-INF';
            $fmtChunkEnd = $chunkEnd !== null ? "$chunkEnd]" : 'INF)';
            $this->output( "  Checking interval $fmtStart, $fmtChunkEnd\n" );
            $this->dfnCheckInterval( $start, $chunkEnd, $batchSize );
 
            $start = $nextStart;
 
        } while ( $nextStart !== false );
    }
 
    private function dfnCheckInterval( $start = null, $end = null, $batchSize = 100 ) {
        $dbw = $this->getDB( DB_PRIMARY );
        $dbr = $this->getDB( DB_REPLICA, [ 'vslow' ] );
 
        $linksTables = [
            // table name => page_id field
            'pagelinks' => 'pl_from',
            'imagelinks' => 'il_from',
            'categorylinks' => 'cl_from',
            'templatelinks' => 'tl_from',
            'externallinks' => 'el_from',
            'iwlinks' => 'iwl_from',
            'langlinks' => 'll_from',
            'redirect' => 'rd_from',
            'page_props' => 'pp_page',
        ];
 
        foreach ( $linksTables as $table => $field ) {
            $this->output( "    $table: 0" );
            $tableStart = $start;
            $counter = 0;
            do {
                $ids = $dbr->newSelectQueryBuilder()
                    ->select( $field )
                    ->distinct()
                    ->from( $table )
                    ->leftJoin( 'page', null, "$field = page_id" )
                    ->where( self::intervalCond( $dbr, $field, $tableStart, $end ) )
                    ->andWhere( [ 'page_id' => null ] )
                    ->orderBy( $field )
                    ->limit( $batchSize )
                    ->caller( __METHOD__ )->fetchFieldValues();
 
                $numIds = count( $ids );
                if ( $numIds ) {
                    $counter += $numIds;
                    $dbw->delete( $table, [ $field => $ids ], __METHOD__ );
                    $this->output( ", $counter" );
                    $tableStart = $ids[$numIds - 1] + 1;
                    $this->waitForReplication();
                }
 
            } while ( $numIds >= $batchSize && ( $end === null || $tableStart <= $end ) );
 
            $this->output( " deleted.\n" );
        }
    }
 
    private static function intervalCond( IReadableDatabase $db, $var, $start, $end ) {
        if ( $start === null && $end === null ) {
            return "$var IS NOT NULL";
        } elseif ( $end === null ) {
            return "$var >= " . $db->addQuotes( $start );
        } elseif ( $start === null ) {
            return "$var <= " . $db->addQuotes( $end );
        } else {
            return "$var BETWEEN " . $db->addQuotes( $start ) . ' AND ' . $db->addQuotes( $end );
        }
    }
 
    private function refreshTrackingCategory( SelectQueryBuilder $builder, $category ) {
        $cats = $this->getPossibleCategories( $category );
 
        if ( !$cats ) {
            $this->error( "Tracking category '$category' is disabled\n" );
            // Output to stderr but don't bail out.
        }
 
        foreach ( $cats as $cat ) {
            $this->refreshCategory( clone $builder, $cat );
        }
    }
 
    private function refreshCategory( SelectQueryBuilder $builder, LinkTarget $category ) {
        $this->output( "Refreshing pages in category '{$category->getText()}'...\n" );
 
        $builder->join( 'categorylinks', null, 'page_id=cl_from' )
            ->andWhere( [ 'cl_to' => $category->getDBkey() ] );
        $this->doRefreshLinks( $builder, false, [ 'cl_timestamp', 'cl_from' ] );
    }
 
    private function getPossibleCategories( $categoryKey ) {
        $cats = $this->getServiceContainer()->getTrackingCategories()->getTrackingCategories();
        if ( isset( $cats[$categoryKey] ) ) {
            return $cats[$categoryKey]['cats'];
        }
        $this->fatalError( "Unknown tracking category {$categoryKey}\n" );
    }
}
 
$maintClass = RefreshLinks::class;
require_once RUN_MAINTENANCE_IF_MAIN;