Cache NS_TEMPLATE/NS_FILE titles in LinkCache persistently
[lhc/web/wiklou.git] / includes / page / WikiPage.php
index 033eb67..f1e59de 100644 (file)
@@ -21,6 +21,7 @@
  */
 
 use \MediaWiki\Logger\LoggerFactory;
+use \MediaWiki\MediaWikiServices;
 
 /**
  * Class representing a MediaWiki article and history.
@@ -134,7 +135,7 @@ class WikiPage implements Page, IDBAccessObject {
         *
         * @param int $id Article ID to load
         * @param string|int $from One of the following values:
-        *        - "fromdb" or WikiPage::READ_NORMAL to select from a slave database
+        *        - "fromdb" or WikiPage::READ_NORMAL to select from a replica DB
         *        - "fromdbmaster" or WikiPage::READ_LATEST to select from the master database
         *
         * @return WikiPage|null
@@ -146,7 +147,7 @@ class WikiPage implements Page, IDBAccessObject {
                }
 
                $from = self::convertSelectType( $from );
-               $db = wfGetDB( $from === self::READ_LATEST ? DB_MASTER : DB_SLAVE );
+               $db = wfGetDB( $from === self::READ_LATEST ? DB_MASTER : DB_REPLICA );
                $row = $db->selectRow(
                        'page', self::selectFields(), [ 'page_id' => $id ], __METHOD__ );
                if ( !$row ) {
@@ -161,7 +162,7 @@ class WikiPage implements Page, IDBAccessObject {
         * @since 1.20
         * @param object $row Database row containing at least fields returned by selectFields().
         * @param string|int $from Source of $data:
-        *        - "fromdb" or WikiPage::READ_NORMAL: from a slave DB
+        *        - "fromdb" or WikiPage::READ_NORMAL: from a replica DB
         *        - "fromdbmaster" or WikiPage::READ_LATEST: from the master DB
         *        - "forupdate" or WikiPage::READ_LOCKING: from the master DB using SELECT FOR UPDATE
         * @return WikiPage
@@ -346,7 +347,7 @@ class WikiPage implements Page, IDBAccessObject {
         *
         * @param object|string|int $from One of the following:
         *   - A DB query result object.
-        *   - "fromdb" or WikiPage::READ_NORMAL to get from a slave DB.
+        *   - "fromdb" or WikiPage::READ_NORMAL to get from a replica DB.
         *   - "fromdbmaster" or WikiPage::READ_LATEST to get from the master DB.
         *   - "forupdate"  or WikiPage::READ_LOCKING to get from the master DB
         *     using SELECT FOR UPDATE.
@@ -365,7 +366,7 @@ class WikiPage implements Page, IDBAccessObject {
                        $data = $this->pageDataFromTitle( wfGetDB( $index ), $this->mTitle, $opts );
 
                        if ( !$data
-                               && $index == DB_SLAVE
+                               && $index == DB_REPLICA
                                && wfGetLB()->getServerCount() > 1
                                && wfGetLB()->hasOrMadeRecentMasterChanges()
                        ) {
@@ -374,7 +375,7 @@ class WikiPage implements Page, IDBAccessObject {
                                $data = $this->pageDataFromTitle( wfGetDB( $index ), $this->mTitle, $opts );
                        }
                } else {
-                       // No idea from where the caller got this data, assume slave database.
+                       // No idea from where the caller got this data, assume replica DB.
                        $data = $from;
                        $from = self::READ_NORMAL;
                }
@@ -388,7 +389,7 @@ class WikiPage implements Page, IDBAccessObject {
         * @since 1.20
         * @param object|bool $data DB row containing fields returned by selectFields() or false
         * @param string|int $from One of the following:
-        *        - "fromdb" or WikiPage::READ_NORMAL if the data comes from a slave DB
+        *        - "fromdb" or WikiPage::READ_NORMAL if the data comes from a replica DB
         *        - "fromdbmaster" or WikiPage::READ_LATEST if the data comes from the master DB
         *        - "forupdate"  or WikiPage::READ_LOCKING if the data comes from
         *          the master DB using SELECT FOR UPDATE
@@ -504,13 +505,13 @@ class WikiPage implements Page, IDBAccessObject {
 
        /**
         * Loads page_touched and returns a value indicating if it should be used
-        * @return bool True if not a redirect
+        * @return bool True if this page exists and is not a redirect
         */
        public function checkTouched() {
                if ( !$this->mDataLoaded ) {
                        $this->loadPageData();
                }
-               return !$this->mIsRedirect;
+               return ( $this->mId && !$this->mIsRedirect );
        }
 
        /**
@@ -552,9 +553,9 @@ class WikiPage implements Page, IDBAccessObject {
         */
        public function getOldestRevision() {
 
-               // Try using the slave database first, then try the master
+               // Try using the replica DB first, then try the master
                $continue = 2;
-               $db = wfGetDB( DB_SLAVE );
+               $db = wfGetDB( DB_REPLICA );
                $revSelectFields = Revision::selectFields();
 
                $row = null;
@@ -609,7 +610,7 @@ class WikiPage implements Page, IDBAccessObject {
                        $flags = Revision::READ_LOCKING;
                } elseif ( $this->mDataLoadedFrom == self::READ_LATEST ) {
                        // Bug T93976: if page_latest was loaded from the master, fetch the
-                       // revision from there as well, as it may not exist yet on a slave DB.
+                       // revision from there as well, as it may not exist yet on a replica DB.
                        // Also, this keeps the queries in the same REPEATABLE-READ snapshot.
                        $flags = Revision::READ_LATEST;
                } else {
@@ -831,7 +832,7 @@ class WikiPage implements Page, IDBAccessObject {
                                // links.
                                $hasLinks = (bool)count( $editInfo->output->getLinks() );
                        } else {
-                               $hasLinks = (bool)wfGetDB( DB_SLAVE )->selectField( 'pagelinks', 1,
+                               $hasLinks = (bool)wfGetDB( DB_REPLICA )->selectField( 'pagelinks', 1,
                                        [ 'pl_from' => $this->getId() ], __METHOD__ );
                        }
                }
@@ -856,7 +857,7 @@ class WikiPage implements Page, IDBAccessObject {
                }
 
                // Query the redirect table
-               $dbr = wfGetDB( DB_SLAVE );
+               $dbr = wfGetDB( DB_REPLICA );
                $row = $dbr->selectRow( 'redirect',
                        [ 'rd_namespace', 'rd_title', 'rd_fragment', 'rd_interwiki' ],
                        [ 'rd_from' => $this->getId() ],
@@ -989,7 +990,7 @@ class WikiPage implements Page, IDBAccessObject {
        public function getContributors() {
                // @todo FIXME: This is expensive; cache this info somewhere.
 
-               $dbr = wfGetDB( DB_SLAVE );
+               $dbr = wfGetDB( DB_REPLICA );
 
                if ( $dbr->implicitGroupby() ) {
                        $realNameField = 'user_real_name';
@@ -1386,7 +1387,7 @@ class WikiPage implements Page, IDBAccessObject {
 
                $baseRevId = null;
                if ( $edittime && $sectionId !== 'new' ) {
-                       $dbr = wfGetDB( DB_SLAVE );
+                       $dbr = wfGetDB( DB_REPLICA );
                        $rev = Revision::loadFromTimestamp( $dbr, $this->mTitle, $edittime );
                        // Try the master if this thread may have just added it.
                        // This could be abstracted into a Revision method, but we don't want
@@ -2110,7 +2111,7 @@ class WikiPage implements Page, IDBAccessObject {
                                // We get here if vary-revision is set. This means that this page references
                                // itself (such as via self-transclusion). In this case, we need to make sure
                                // that any such self-references refer to the newly-saved revision, and not
-                               // to the previous one, which could otherwise happen due to slave lag.
+                               // to the previous one, which could otherwise happen due to replica DB lag.
                                $oldCallback = $edit->popts->getCurrentRevisionCallback();
                                $edit->popts->setCurrentRevisionCallback(
                                        function ( Title $title, $parser = false ) use ( $revision, &$oldCallback ) {
@@ -2713,14 +2714,14 @@ class WikiPage implements Page, IDBAccessObject {
                $protectDescription = '';
 
                foreach ( array_filter( $limit ) as $action => $restrictions ) {
-                       # $action is one of $wgRestrictionTypes = array( 'create', 'edit', 'move', 'upload' ).
+                       # $action is one of $wgRestrictionTypes = [ 'create', 'edit', 'move', 'upload' ].
                        # All possible message keys are listed here for easier grepping:
                        # * restriction-create
                        # * restriction-edit
                        # * restriction-move
                        # * restriction-upload
                        $actionText = wfMessage( 'restriction-' . $action )->inContentLanguage()->text();
-                       # $restrictions is one of $wgRestrictionLevels = array( '', 'autoconfirmed', 'sysop' ),
+                       # $restrictions is one of $wgRestrictionLevels = [ '', 'autoconfirmed', 'sysop' ],
                        # with '' filtered out. All possible message keys are listed below:
                        # * protect-level-autoconfirmed
                        # * protect-level-sysop
@@ -2873,13 +2874,24 @@ class WikiPage implements Page, IDBAccessObject {
                        return $status;
                }
 
+               // Given the lock above, we can be confident in the title and page ID values
+               $namespace = $this->getTitle()->getNamespace();
+               $dbKey = $this->getTitle()->getDBkey();
+
                // At this point we are now comitted to returning an OK
                // status unless some DB query error or other exception comes up.
                // This way callers don't have to call rollback() if $status is bad
                // unless they actually try to catch exceptions (which is rare).
 
                // we need to remember the old content so we can use it to generate all deletion updates.
-               $content = $this->getContent( Revision::RAW );
+               try {
+                       $content = $this->getContent( Revision::RAW );
+               } catch ( Exception $ex ) {
+                       wfLogWarning( __METHOD__ . ': failed to load content during deletion! '
+                               . $ex->getMessage() );
+
+                       $content = null;
+               }
 
                // Bitfields to further suppress the content
                if ( $suppress ) {
@@ -2889,58 +2901,56 @@ class WikiPage implements Page, IDBAccessObject {
                        $bitfield |= Revision::DELETED_COMMENT;
                        $bitfield |= Revision::DELETED_USER;
                        $bitfield |= Revision::DELETED_RESTRICTED;
+                       $deletionFields = [ $dbw->addQuotes( $bitfield ) . ' AS deleted' ];
                } else {
-                       $bitfield = 'rev_deleted';
-               }
-
-               /**
-                * For now, shunt the revision data into the archive table.
-                * Text is *not* removed from the text table; bulk storage
-                * is left intact to avoid breaking block-compression or
-                * immutable storage schemes.
-                *
-                * For backwards compatibility, note that some older archive
-                * table entries will have ar_text and ar_flags fields still.
-                *
-                * In the future, we may keep revisions and mark them with
-                * the rev_deleted field, which is reserved for this purpose.
-                */
-
-               $row = [
-                       'ar_namespace'  => 'page_namespace',
-                       'ar_title'      => 'page_title',
-                       'ar_comment'    => 'rev_comment',
-                       'ar_user'       => 'rev_user',
-                       'ar_user_text'  => 'rev_user_text',
-                       'ar_timestamp'  => 'rev_timestamp',
-                       'ar_minor_edit' => 'rev_minor_edit',
-                       'ar_rev_id'     => 'rev_id',
-                       'ar_parent_id'  => 'rev_parent_id',
-                       'ar_text_id'    => 'rev_text_id',
-                       'ar_text'       => '\'\'', // Be explicit to appease
-                       'ar_flags'      => '\'\'', // MySQL's "strict mode"...
-                       'ar_len'        => 'rev_len',
-                       'ar_page_id'    => 'page_id',
-                       'ar_deleted'    => $bitfield,
-                       'ar_sha1'       => 'rev_sha1',
-               ];
-
-               if ( $wgContentHandlerUseDB ) {
-                       $row['ar_content_model'] = 'rev_content_model';
-                       $row['ar_content_format'] = 'rev_content_format';
-               }
-
-               // Copy all the page revisions into the archive table
-               $dbw->insertSelect(
-                       'archive',
-                       [ 'page', 'revision' ],
-                       $row,
-                       [
-                               'page_id' => $id,
-                               'page_id = rev_page'
-                       ],
-                       __METHOD__
+                       $deletionFields = [ 'rev_deleted AS deleted' ];
+               }
+
+               // For now, shunt the revision data into the archive table.
+               // Text is *not* removed from the text table; bulk storage
+               // is left intact to avoid breaking block-compression or
+               // immutable storage schemes.
+               // In the future, we may keep revisions and mark them with
+               // the rev_deleted field, which is reserved for this purpose.
+
+               // Get all of the page revisions
+               $fields = array_diff( Revision::selectFields(), [ 'rev_deleted' ] );
+               $res = $dbw->select(
+                       'revision',
+                       array_merge( $fields, $deletionFields ),
+                       [ 'rev_page' => $id ],
+                       __METHOD__,
+                       'FOR UPDATE'
                );
+               // Build their equivalent archive rows
+               $rowsInsert = [];
+               foreach ( $res as $row ) {
+                       $rowInsert = [
+                               'ar_namespace'  => $namespace,
+                               'ar_title'      => $dbKey,
+                               'ar_comment'    => $row->rev_comment,
+                               'ar_user'       => $row->rev_user,
+                               'ar_user_text'  => $row->rev_user_text,
+                               'ar_timestamp'  => $row->rev_timestamp,
+                               'ar_minor_edit' => $row->rev_minor_edit,
+                               'ar_rev_id'     => $row->rev_id,
+                               'ar_parent_id'  => $row->rev_parent_id,
+                               'ar_text_id'    => $row->rev_text_id,
+                               'ar_text'       => '',
+                               'ar_flags'      => '',
+                               'ar_len'        => $row->rev_len,
+                               'ar_page_id'    => $id,
+                               'ar_deleted'    => $row->deleted,
+                               'ar_sha1'       => $row->rev_sha1,
+                       ];
+                       if ( $wgContentHandlerUseDB ) {
+                               $rowInsert['ar_content_model'] = $row->rev_content_model;
+                               $rowInsert['ar_content_format'] = $row->rev_content_format;
+                       }
+                       $rowsInsert[] = $rowInsert;
+               }
+               // Copy them into the archive table
+               $dbw->insert( 'archive', $rowsInsert, __METHOD__ );
                // Save this so we can pass it to the ArticleDeleteComplete hook.
                $archivedRevisionCount = $dbw->affectedRows();
 
@@ -3024,8 +3034,16 @@ class WikiPage implements Page, IDBAccessObject {
         *   may already return null when the page proper was deleted.
         */
        public function doDeleteUpdates( $id, Content $content = null ) {
+               try {
+                       $countable = $this->isCountable();
+               } catch ( Exception $ex ) {
+                       // fallback for deleting broken pages for which we cannot load the content for
+                       // some reason. Note that doDeleteArticleReal() already logged this problem.
+                       $countable = false;
+               }
+
                // Update site status
-               DeferredUpdates::addUpdate( new SiteStatsUpdate( 0, 1, - (int)$this->isCountable(), -1 ) );
+               DeferredUpdates::addUpdate( new SiteStatsUpdate( 0, 1, - (int)$countable, -1 ) );
 
                // Delete pagelinks, update secondary indexes, etc
                $updates = $this->getDeletionUpdates( $content );
@@ -3304,9 +3322,11 @@ class WikiPage implements Page, IDBAccessObject {
                $title->purgeSquid();
                $title->deleteTitleProtection();
 
+               MediaWikiServices::getInstance()->getLinkCache()->invalidateTitle( $title );
+
                if ( $title->getNamespace() == NS_CATEGORY ) {
                        // Load the Category object, which will schedule a job to create
-                       // the category table row if necessary. Checking a slave is ok
+                       // the category table row if necessary. Checking a replica DB is ok
                        // here, in the worst case it'll run an unnecessary recount job on
                        // a category that probably doesn't have many members.
                        Category::newFromTitle( $title )->getID();
@@ -3329,6 +3349,8 @@ class WikiPage implements Page, IDBAccessObject {
                $title->touchLinks();
                $title->purgeSquid();
 
+               MediaWikiServices::getInstance()->getLinkCache()->invalidateTitle( $title );
+
                // File cache
                HTMLFileCache::clearFileCache( $title );
                InfoAction::invalidateCache( $title );
@@ -3372,6 +3394,8 @@ class WikiPage implements Page, IDBAccessObject {
                // Invalidate the caches of all pages which redirect here
                DeferredUpdates::addUpdate( new HTMLCacheUpdate( $title, 'redirect' ) );
 
+               MediaWikiServices::getInstance()->getLinkCache()->invalidateTitle( $title );
+
                // Purge CDN for this page only
                $title->purgeSquid();
                // Clear file cache for this page only
@@ -3397,7 +3421,7 @@ class WikiPage implements Page, IDBAccessObject {
                        return TitleArray::newFromResult( new FakeResultWrapper( [] ) );
                }
 
-               $dbr = wfGetDB( DB_SLAVE );
+               $dbr = wfGetDB( DB_REPLICA );
                $res = $dbr->select( 'categorylinks',
                        [ 'cl_to AS page_title, ' . NS_CATEGORY . ' AS page_namespace' ],
                        // Have to do that since DatabaseBase::fieldNamesWithAlias treats numeric indexes
@@ -3422,7 +3446,7 @@ class WikiPage implements Page, IDBAccessObject {
                        return [];
                }
 
-               $dbr = wfGetDB( DB_SLAVE );
+               $dbr = wfGetDB( DB_REPLICA );
                $res = $dbr->select( [ 'categorylinks', 'page_props', 'page' ],
                        [ 'cl_to' ],
                        [ 'cl_from' => $id, 'pp_page=page_id', 'pp_propname' => 'hiddencat',
@@ -3641,7 +3665,14 @@ class WikiPage implements Page, IDBAccessObject {
                if ( !$content ) {
                        // load content object, which may be used to determine the necessary updates.
                        // XXX: the content may not be needed to determine the updates.
-                       $content = $this->getContent( Revision::RAW );
+                       try {
+                               $content = $this->getContent( Revision::RAW );
+                       } catch ( Exception $ex ) {
+                               // If we can't load the content, something is wrong. Perhaps that's why
+                               // the user is trying to delete the page, so let's not fail in that case.
+                               // Note that doDeleteArticleReal() will already have logged an issue with
+                               // loading the content.
+                       }
                }
 
                if ( !$content ) {