refreshLinks.php

Go to the documentation of this file.

<?php
use MediaWiki\MediaWikiServices;
use MediaWiki\Revision\RevisionRecord;
use Wikimedia\Rdbms\IDatabase;
 
require_once __DIR__ . '/Maintenance.php';
 
class RefreshLinks extends Maintenance {
    private const REPORTING_INTERVAL = 100;
 
    protected $namespace = false;
 
    public function __construct() {
        parent::__construct();
        $this->addDescription( 'Refresh link tables' );
        $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( '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->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 );
 
        $ns = $this->getOption( 'namespace' );
        if ( $ns === null ) {
            $this->namespace = false;
        } else {
            $this->namespace = (int)$ns;
        }
 
        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( $title );
        } elseif ( $this->hasOption( 'tracking-category' ) ) {
            $this->refreshTrackingCategory( $this->getOption( 'trackingcategory' ) );
        } elseif ( !$this->hasOption( 'dfn-only' ) ) {
            $new = $this->hasOption( 'new-only' );
            $redir = $this->hasOption( 'redirects-only' );
            $oldRedir = $this->hasOption( 'old-redirects-only' );
            $this->doRefreshLinks( $start, $new, $end, $redir, $oldRedir );
            $this->deleteLinksFromNonexistent( null, null, $this->getBatchSize(), $dfnChunkSize );
        } else {
            $this->deleteLinksFromNonexistent( $start, $end, $this->getBatchSize(), $dfnChunkSize );
        }
    }
 
    private function namespaceCond() {
        return $this->namespace !== false
            ? [ 'page_namespace' => $this->namespace ]
            : [];
    }
 
    private function doRefreshLinks( $start, $newOnly = false,
        $end = null, $redirectsOnly = false, $oldRedirectsOnly = false
    ) {
        $dbr = $this->getDB( DB_REPLICA, [ 'vslow' ] );
 
        if ( $start === null ) {
            $start = 1;
        }
 
        // Give extensions a chance to optimize settings
        $this->getHookRunner()->onMaintenanceRefreshLinksInit( $this );
 
        $what = $redirectsOnly ? "redirects" : "links";
        $lbFactory = MediaWikiServices::getInstance()->getDBLoadBalancerFactory();
 
        if ( $oldRedirectsOnly ) {
            # This entire code path is cut-and-pasted from below.  Hurrah.
 
            $conds = [
                "page_is_redirect=1",
                "rd_from IS NULL",
                self::intervalCond( $dbr, 'page_id', $start, $end ),
            ] + $this->namespaceCond();
 
            $res = $dbr->select(
                [ 'page', 'redirect' ],
                'page_id',
                $conds,
                __METHOD__,
                [],
                [ 'redirect' => [ "LEFT JOIN", "page_id=rd_from" ] ]
            );
            $num = $res->numRows();
            $this->output( "Refreshing $num old redirects from $start...\n" );
 
            $i = 0;
 
            foreach ( $res as $row ) {
                if ( !( ++$i % self::REPORTING_INTERVAL ) ) {
                    $this->output( "$i\n" );
                    $lbFactory->waitForReplication();
                }
                $this->fixRedirect( $row->page_id );
            }
        } elseif ( $newOnly ) {
            $this->output( "Refreshing $what from " );
            $res = $dbr->select( 'page',
                [ 'page_id' ],
                [
                    'page_is_new' => 1,
                    self::intervalCond( $dbr, 'page_id', $start, $end ),
                ] + $this->namespaceCond(),
                __METHOD__
            );
            $num = $res->numRows();
            $this->output( "$num new articles...\n" );
 
            $i = 0;
            foreach ( $res as $row ) {
                if ( !( ++$i % self::REPORTING_INTERVAL ) ) {
                    $this->output( "$i\n" );
                    $lbFactory->waitForReplication();
                }
                if ( $redirectsOnly ) {
                    $this->fixRedirect( $row->page_id );
                } else {
                    self::fixLinksFromArticle( $row->page_id, $this->namespace );
                }
            }
        } else {
            if ( !$end ) {
                $maxPage = $dbr->selectField( 'page', 'max(page_id)', '', __METHOD__ );
                $maxRD = $dbr->selectField( 'redirect', 'max(rd_from)', '', __METHOD__ );
                $end = max( $maxPage, $maxRD );
            }
            $this->output( "Refreshing redirects table.\n" );
            $this->output( "Starting from page_id $start of $end.\n" );
 
            for ( $id = $start; $id <= $end; $id++ ) {
                if ( !( $id % self::REPORTING_INTERVAL ) ) {
                    $this->output( "$id\n" );
                    $lbFactory->waitForReplication();
                }
                $this->fixRedirect( $id );
            }
 
            if ( !$redirectsOnly ) {
                $this->output( "Refreshing links tables.\n" );
                $this->output( "Starting from page_id $start of $end.\n" );
 
                for ( $id = $start; $id <= $end; $id++ ) {
                    if ( !( $id % self::REPORTING_INTERVAL ) ) {
                        $this->output( "$id\n" );
                        $lbFactory->waitForReplication();
                    }
                    self::fixLinksFromArticle( $id, $this->namespace );
                }
            }
        }
    }
 
    private function fixRedirect( $id ) {
        $page = MediaWikiServices::getInstance()->getWikiPageFactory()->newFromID( $id );
        $dbw = $this->getDB( DB_PRIMARY );
 
        if ( $page === null ) {
            // This page doesn't exist (any more)
            // Delete any redirect table entry for it
            $dbw->delete( 'redirect', [ 'rd_from' => $id ],
                __METHOD__ );
 
            return;
        } elseif ( $this->namespace !== false
            && !$page->getTitle()->inNamespace( $this->namespace )
        ) {
            return;
        }
 
        $rt = null;
        $content = $page->getContent( RevisionRecord::RAW );
        if ( $content !== null ) {
            $rt = $content->getUltimateRedirectTarget();
        }
 
        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, $ns = false ) {
        $services = MediaWikiServices::getInstance();
        $page = $services->getWikiPageFactory()->newFromID( $id );
 
        $services->getLinkCache()->clear();
 
        if ( $page === null ) {
            return;
        } elseif ( $ns !== false
            && !$page->getTitle()->inNamespace( $ns ) ) {
            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,
            'recursive' => false,
        ] );
        DeferredUpdates::doUpdates();
    }
 
    private function deleteLinksFromNonexistent( $start = null, $end = null, $batchSize = 100,
        $chunkSize = 100000
    ) {
        MediaWikiServices::getInstance()->getDBLoadBalancerFactory()->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->selectField(
                'page',
                'page_id',
                [ self::intervalCond( $dbr, 'page_id', $start, $end ) ]
                + $this->namespaceCond(),
                __METHOD__,
                [ 'ORDER BY' => 'page_id', 'OFFSET' => $chunkSize ]
            );
 
            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',
        ];
 
        $lbFactory = MediaWikiServices::getInstance()->getDBLoadBalancerFactory();
        foreach ( $linksTables as $table => $field ) {
            $this->output( "    $table: 0" );
            $tableStart = $start;
            $counter = 0;
            do {
                $ids = $dbr->selectFieldValues(
                    $table,
                    $field,
                    [
                        self::intervalCond( $dbr, $field, $tableStart, $end ),
                        "$field NOT IN ({$dbr->selectSQLText( 'page', 'page_id', [], __METHOD__ )})",
                    ],
                    __METHOD__,
                    [ 'DISTINCT', 'ORDER BY' => $field, 'LIMIT' => $batchSize ]
                );
 
                $numIds = count( $ids );
                if ( $numIds ) {
                    $counter += $numIds;
                    $dbw->delete( $table, [ $field => $ids ], __METHOD__ );
                    $this->output( ", $counter" );
                    $tableStart = $ids[$numIds - 1] + 1;
                    $lbFactory->waitForReplication();
                }
 
            } while ( $numIds >= $batchSize && ( $end === null || $tableStart <= $end ) );
 
            $this->output( " deleted.\n" );
        }
    }
 
    private static function intervalCond( IDatabase $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( $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( $cat );
        }
    }
 
    private function refreshCategory( Title $category ) {
        $this->output( "Refreshing pages in category '{$category->getText()}'...\n" );
 
        $dbr = $this->getDB( DB_REPLICA );
        $conds = [
            'page_id=cl_from',
            'cl_to' => $category->getDBkey(),
        ];
        if ( $this->namespace !== false ) {
            $conds['page_namespace'] = $this->namespace;
        }
 
        $i = 0;
        $timestamp = '';
        $lastId = 0;
        $lbFactory = MediaWikiServices::getInstance()->getDBLoadBalancerFactory();
        do {
            $finalConds = $conds;
            $timestamp = $dbr->addQuotes( $timestamp );
            $finalConds[] =
                "(cl_timestamp > $timestamp OR (cl_timestamp = $timestamp AND cl_from > $lastId))";
            $res = $dbr->select( [ 'page', 'categorylinks' ],
                [ 'page_id', 'cl_timestamp' ],
                $finalConds,
                __METHOD__,
                [
                    'ORDER BY' => [ 'cl_timestamp', 'cl_from' ],
                    'LIMIT' => $this->getBatchSize(),
                ]
            );
 
            foreach ( $res as $row ) {
                if ( !( ++$i % self::REPORTING_INTERVAL ) ) {
                    $this->output( "$i\n" );
                    $lbFactory->waitForReplication();
                }
                $lastId = $row->page_id;
                $timestamp = $row->cl_timestamp;
                self::fixLinksFromArticle( $row->page_id );
            }
 
        } while ( $res->numRows() == $this->getBatchSize() );
    }
 
    private function getPossibleCategories( $categoryKey ) {
        $trackingCategories = new TrackingCategories( $this->getConfig() );
        $cats = $trackingCategories->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;