<?php
/**
+ * Refresh link tables.
+ *
* This program is free software; you can redistribute it and/or modify
* it under the terms of the GNU General Public License as published by
* the Free Software Foundation; either version 2 of the License, or
* 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
* http://www.gnu.org/copyleft/gpl.html
*
+ * @file
* @ingroup Maintenance
*/
-require_once( dirname( __FILE__ ) . '/Maintenance.php' );
+use Wikimedia\Rdbms\IDatabase;
+
+require_once __DIR__ . '/Maintenance.php';
+/**
+ * Maintenance script to refresh link tables.
+ *
+ * @ingroup Maintenance
+ */
class RefreshLinks extends Maintenance {
+ const REPORTING_INTERVAL = 100;
+
+ /** @var int|bool */
+ protected $namespace = false;
+
public function __construct() {
parent::__construct();
- $this->mDescription = "Refresh link tables";
+ $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( 'm', 'Maximum replication lag', false, true );
$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() {
- $max = $this->getOption( 'm', 0 );
- if ( !$this->hasOption( 'dfn-only' ) ) {
- $start = $this->getArg( 0, 1 );
- $new = $this->getOption( 'new-only', false );
- $end = $this->getOption( 'e', 0 );
- $redir = $this->getOption( 'redirects-only', false );
- $oldRedir = $this->getOption( 'old-redirects-only', false );
- $this->doRefreshLinks( $start, $new, $max, $end, $redir, $oldRedir );
+ // 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;
}
- $this->deleteLinksFromNonexistent( $max, $this->mBatchSize );
+ if ( ( $category = $this->getOption( 'category', false ) ) !== false ) {
+ $title = Title::makeTitleSafe( NS_CATEGORY, $category );
+ if ( !$title ) {
+ $this->error( "'$category' is an invalid category name!\n", true );
+ }
+ $this->refreshCategory( $title );
+ } elseif ( ( $category = $this->getOption( 'tracking-category', false ) ) !== false ) {
+ $this->refreshTrackingCategory( $category );
+ } elseif ( !$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->mBatchSize, $dfnChunkSize );
+ } else {
+ $this->deleteLinksFromNonexistent( $start, $end, $this->mBatchSize, $dfnChunkSize );
+ }
+ }
+
+ private function namespaceCond() {
+ return $this->namespace !== false
+ ? [ 'page_namespace' => $this->namespace ]
+ : [];
}
/**
* Do the actual link refreshing.
- * @param $start int Page_id to start from
- * @param $newOnly bool Only do pages with 1 edit
- * @param $maxLag int Max DB replication lag
- * @param $end int Page_id to stop at
- * @param $redirectsOnly bool Only fix redirects
- * @param $oldRedirectsOnly bool Only fix redirects without redirect entries
+ * @param int|null $start Page_id to start from
+ * @param bool $newOnly Only do pages with 1 edit
+ * @param int|null $end Page_id to stop at
+ * @param bool $redirectsOnly Only fix redirects
+ * @param bool $oldRedirectsOnly Only fix redirects without redirect entries
*/
- private function doRefreshLinks( $start, $newOnly = false, $maxLag = false,
- $end = 0, $redirectsOnly = false, $oldRedirectsOnly = false ) {
- global $wgUser, $wgParser, $wgUseTidy;
-
- $reportingInterval = 100;
- $dbr = wfGetDB( DB_SLAVE );
- $start = intval( $start );
-
- # Don't generate TeX PNGs (lack of a sensible current directory causes errors anyway)
- $wgUser->setOption( 'math', MW_MATH_SOURCE );
+ private function doRefreshLinks( $start, $newOnly = false,
+ $end = null, $redirectsOnly = false, $oldRedirectsOnly = false
+ ) {
+ $dbr = $this->getDB( DB_REPLICA, [ 'vslow' ] );
- # Don't generate extension images (e.g. Timeline)
- if ( method_exists( $wgParser, "clearTagHooks" ) ) {
- $wgParser->clearTagHooks();
+ if ( $start === null ) {
+ $start = 1;
}
- # Don't use HTML tidy
- $wgUseTidy = false;
+ // Give extensions a chance to optimize settings
+ Hooks::run( 'MaintenanceRefreshLinksInit', [ $this ] );
$what = $redirectsOnly ? "redirects" : "links";
if ( $oldRedirectsOnly ) {
# This entire code path is cut-and-pasted from below. Hurrah.
- $res = $dbr->query(
- "SELECT page_id " .
- "FROM page " .
- "LEFT JOIN redirect ON page_id=rd_from " .
- "WHERE page_is_redirect=1 AND rd_from IS NULL AND " .
- ( $end == 0 ? "page_id >= $start"
- : "page_id BETWEEN $start AND $end" ),
- __METHOD__
+
+ $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 = $dbr->numRows( $res );
+ $num = $res->numRows();
$this->output( "Refreshing $num old redirects from $start...\n" );
$i = 0;
+
foreach ( $res as $row ) {
- if ( !( ++$i % $reportingInterval ) ) {
+ if ( !( ++$i % self::REPORTING_INTERVAL ) ) {
$this->output( "$i\n" );
- wfWaitForSlaves( $maxLag );
+ wfWaitForSlaves();
}
$this->fixRedirect( $row->page_id );
}
} elseif ( $newOnly ) {
$this->output( "Refreshing $what from " );
$res = $dbr->select( 'page',
- array( 'page_id' ),
- array(
+ [ 'page_id' ],
+ [
'page_is_new' => 1,
- "page_id >= $start" ),
+ self::intervalCond( $dbr, 'page_id', $start, $end ),
+ ] + $this->namespaceCond(),
__METHOD__
);
- $num = $dbr->numRows( $res );
+ $num = $res->numRows();
$this->output( "$num new articles...\n" );
$i = 0;
foreach ( $res as $row ) {
- if ( !( ++$i % $reportingInterval ) ) {
+ if ( !( ++$i % self::REPORTING_INTERVAL ) ) {
$this->output( "$i\n" );
- wfWaitForSlaves( $maxLag );
+ wfWaitForSlaves();
}
- if ( $redirectsOnly )
+ if ( $redirectsOnly ) {
$this->fixRedirect( $row->page_id );
- else
- self::fixLinksFromArticle( $row->page_id );
+ } else {
+ self::fixLinksFromArticle( $row->page_id, $this->namespace );
+ }
}
} else {
if ( !$end ) {
$this->output( "Starting from page_id $start of $end.\n" );
for ( $id = $start; $id <= $end; $id++ ) {
-
- if ( !( $id % $reportingInterval ) ) {
+ if ( !( $id % self::REPORTING_INTERVAL ) ) {
$this->output( "$id\n" );
- wfWaitForSlaves( $maxLag );
+ wfWaitForSlaves();
}
$this->fixRedirect( $id );
}
if ( !$redirectsOnly ) {
- $this->output( "Refreshing links table.\n" );
+ $this->output( "Refreshing links tables.\n" );
$this->output( "Starting from page_id $start of $end.\n" );
for ( $id = $start; $id <= $end; $id++ ) {
-
- if ( !( $id % $reportingInterval ) ) {
+ if ( !( $id % self::REPORTING_INTERVAL ) ) {
$this->output( "$id\n" );
- wfWaitForSlaves( $maxLag );
+ wfWaitForSlaves();
}
- self::fixLinksFromArticle( $id );
+ self::fixLinksFromArticle( $id, $this->namespace );
}
}
}
}
/**
- * Update the redirect entry for a given page
- * @param $id int The page_id of the redirect
+ * Update the redirect entry for a given page.
+ *
+ * This methods bypasses the "redirect" table to get the redirect target,
+ * and parses the page's content to fetch it. This allows to be sure that
+ * the redirect target is up to date and valid.
+ * This is particularly useful when modifying namespaces to be sure the
+ * entry in the "redirect" table points to the correct page and not to an
+ * invalid one.
+ *
+ * @param int $id The page ID to check
*/
private function fixRedirect( $id ) {
- global $wgTitle, $wgArticle;
+ $page = WikiPage::newFromID( $id );
+ $dbw = $this->getDB( DB_MASTER );
- $wgTitle = Title::newFromID( $id );
- $dbw = wfGetDB( DB_MASTER );
-
- if ( is_null( $wgTitle ) ) {
+ if ( $page === null ) {
// This page doesn't exist (any more)
// Delete any redirect table entry for it
- $dbw->delete( 'redirect', array( 'rd_from' => $id ),
+ $dbw->delete( 'redirect', [ 'rd_from' => $id ],
__METHOD__ );
+
+ return;
+ } elseif ( $this->namespace !== false
+ && !$page->getTitle()->inNamespace( $this->namespace )
+ ) {
return;
}
- $wgArticle = new Article( $wgTitle );
- $rt = $wgArticle->followRedirect();
+ $rt = null;
+ $content = $page->getContent( Revision::RAW );
+ if ( $content !== null ) {
+ $rt = $content->getUltimateRedirectTarget();
+ }
- if ( !$rt || !is_object( $rt ) ) {
- // $wgTitle is not a redirect
+ if ( $rt === null ) {
+ // The page is not a redirect
// Delete any redirect table entry for it
- $dbw->delete( 'redirect', array( 'rd_from' => $id ),
- __METHOD__ );
+ $dbw->delete( 'redirect', [ 'rd_from' => $id ], __METHOD__ );
+ $fieldValue = 0;
} else {
- $wgArticle->updateRedirectOn( $dbw, $rt );
+ $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__ );
}
/**
* Run LinksUpdate for all links on a given page_id
- * @param $id int The page_id
+ * @param int $id The page_id
+ * @param int|bool $ns Only fix links if it is in this namespace
*/
- public static function fixLinksFromArticle( $id ) {
- global $wgTitle, $wgParser;
-
- $wgTitle = Title::newFromID( $id );
- $dbw = wfGetDB( DB_MASTER );
+ public static function fixLinksFromArticle( $id, $ns = false ) {
+ $page = WikiPage::newFromID( $id );
LinkCache::singleton()->clear();
- if ( is_null( $wgTitle ) ) {
+ if ( $page === null ) {
+ return;
+ } elseif ( $ns !== false
+ && !$page->getTitle()->inNamespace( $ns ) ) {
return;
}
- $dbw->begin();
- $revision = Revision::newFromTitle( $wgTitle );
- if ( !$revision ) {
+ $content = $page->getContent( Revision::RAW );
+ if ( $content === null ) {
return;
}
- $options = new ParserOptions;
- $parserOutput = $wgParser->parse( $revision->getText(), $wgTitle, $options, true, true, $revision->getId() );
- $update = new LinksUpdate( $wgTitle, $parserOutput, false );
- $update->doUpdate();
- $dbw->commit();
+ $updates = $content->getSecondaryDataUpdates(
+ $page->getTitle(), /* $old = */ null, /* $recursive = */ false );
+ foreach ( $updates as $update ) {
+ DeferredUpdates::addUpdate( $update );
+ DeferredUpdates::doUpdates();
+ }
}
- /*
+ /**
* Removes non-existing links from pages from pagelinks, imagelinks,
- * categorylinks, templatelinks and externallinks tables.
+ * categorylinks, templatelinks, externallinks, interwikilinks, langlinks and redirect tables.
*
- * @param $maxLag
- * @param $batchSize The size of deletion batches
+ * @param int|null $start Page_id to start from
+ * @param int|null $end Page_id to stop at
+ * @param int $batchSize The size of deletion batches
+ * @param int $chunkSize Maximum number of existent IDs to check per query
*
* @author Merlijn van Deen <valhallasw@arctus.nl>
*/
- private function deleteLinksFromNonexistent( $maxLag = 0, $batchSize = 100 ) {
- wfWaitForSlaves( $maxLag );
+ private function deleteLinksFromNonexistent( $start = null, $end = null, $batchSize = 100,
+ $chunkSize = 100000
+ ) {
+ wfWaitForSlaves();
+ $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 );
- $dbw = wfGetDB( DB_MASTER );
+ $start = $nextStart;
- $lb = wfGetLBFactory()->newMainLB();
- $dbr = $lb->getConnection( DB_SLAVE );
- $dbr->bufferResults( false );
+ } while ( $nextStart !== false );
+ }
- $linksTables = array( // table name => page_id field
+ /**
+ * @see RefreshLinks::deleteLinksFromNonexistent()
+ * @param int|null $start Page_id to start from
+ * @param int|null $end Page_id to stop at
+ * @param int $batchSize The size of deletion batches
+ */
+ private function dfnCheckInterval( $start = null, $end = null, $batchSize = 100 ) {
+ $dbw = $this->getDB( DB_MASTER );
+ $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( "Retrieving illegal entries from $table... " );
-
- // SELECT DISTINCT( $field ) FROM $table LEFT JOIN page ON $field=page_id WHERE page_id IS NULL;
- $results = $dbr->select( array( $table, 'page' ),
- $field,
- array( 'page_id' => null ),
- __METHOD__,
- 'DISTINCT',
- array( 'page' => array( 'LEFT JOIN', "$field=page_id" ) )
+ $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__,
+ [ '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;
+ wfWaitForSlaves();
+ }
+
+ } while ( $numIds >= $batchSize && ( $end === null || $tableStart <= $end ) );
+
+ $this->output( " deleted.\n" );
+ }
+ }
+
+ /**
+ * Build a SQL expression for a closed interval (i.e. BETWEEN).
+ *
+ * By specifying a null $start or $end, it is also possible to create
+ * half-bounded or unbounded intervals using this function.
+ *
+ * @param IDatabase $db
+ * @param string $var Field name
+ * @param mixed $start First value to include or null
+ * @param mixed $end Last value to include or null
+ * @return string
+ */
+ 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 )}";
+ }
+ }
+
+ /**
+ * Refershes links for pages in a tracking category
+ *
+ * @param string $category Category key
+ */
+ 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 );
+ }
+ }
+
+ /**
+ * Refreshes links to a category
+ *
+ * @param Title $category
+ */
+ 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;
+ 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->mBatchSize,
+ ]
);
- $counter = 0;
- $list = array();
- $this->output( "0.." );
-
- foreach ( $results as $row ) {
- $counter++;
- $list[] = $row->$field;
- if ( ( $counter % $batchSize ) == 0 ) {
- wfWaitForSlaves( 5 );
- $dbw->delete( $table, array( $field => $list ), __METHOD__ );
-
- $this->output( $counter . ".." );
- $list = array();
+ foreach ( $res as $row ) {
+ if ( !( ++$i % self::REPORTING_INTERVAL ) ) {
+ $this->output( "$i\n" );
+ wfWaitForSlaves();
}
+ $lastId = $row->page_id;
+ $timestamp = $row->cl_timestamp;
+ self::fixLinksFromArticle( $row->page_id );
}
- $this->output( $counter );
- if ( count( $list ) > 0 ) {
- $dbw->delete( $table, array( $field => $list ), __METHOD__ );
- }
- $this->output( "\n" );
+
+ } while ( $res->numRows() == $this->mBatchSize );
+ }
+
+ /**
+ * Returns a list of possible categories for a given tracking category key
+ *
+ * @param string $categoryKey
+ * @return Title[]
+ */
+ private function getPossibleCategories( $categoryKey ) {
+ $trackingCategories = new TrackingCategories( $this->getConfig() );
+ $cats = $trackingCategories->getTrackingCategories();
+ if ( isset( $cats[$categoryKey] ) ) {
+ return $cats[$categoryKey]['cats'];
}
- $lb->closeAll();
+ $this->error( "Unknown tracking category {$categoryKey}\n", true );
}
}
$maintClass = 'RefreshLinks';
-require_once( RUN_MAINTENANCE_IF_MAIN );
+require_once RUN_MAINTENANCE_IF_MAIN;