From 183851b06bd6c52f3cae5375f433da720d410447 Mon Sep 17 00:00:00 2001 From: Pierre Schmitz Date: Wed, 11 Oct 2006 18:12:39 +0000 Subject: MediaWiki 1.7.1 wiederhergestellt --- maintenance/storage/compressOld.inc | 300 ++++++++++++++++++++++++++++++++++++ 1 file changed, 300 insertions(+) create mode 100644 maintenance/storage/compressOld.inc (limited to 'maintenance/storage/compressOld.inc') diff --git a/maintenance/storage/compressOld.inc b/maintenance/storage/compressOld.inc new file mode 100644 index 00000000..b7d7094f --- /dev/null +++ b/maintenance/storage/compressOld.inc @@ -0,0 +1,300 @@ +select( 'text', array( 'old_id','old_flags','old_namespace','old_title','old_text' ), + "old_id>=$start", $fname, array( 'ORDER BY' => 'old_id', 'LIMIT' => $chunksize, 'FOR UPDATE' ) ); + if( $dbw->numRows( $res ) == 0 ) { + break; + } + $last = $start; + while( $row = $dbw->fetchObject( $res ) ) { + # print " {$row->old_id} - {$row->old_namespace}:{$row->old_title}\n"; + compressPage( $row, $extdb ); + $last = $row->old_id; + } + $dbw->freeResult( $res ); + $start = $last + 1; # Deletion may leave long empty stretches + print "$start...\n"; + } while( true ); +} + +/** @todo document */ +function compressPage( $row, $extdb ) { + $fname = 'compressPage'; + if ( false !== strpos( $row->old_flags, 'gzip' ) || false !== strpos( $row->old_flags, 'object' ) ) { + #print "Already compressed row {$row->old_id}\n"; + return false; + } + $dbw =& wfGetDB( DB_MASTER ); + $flags = $row->old_flags ? "{$row->old_flags},gzip" : "gzip"; + $compress = gzdeflate( $row->old_text ); + + # Store in external storage if required + if ( $extdb !== '' ) { + $storeObj = new ExternalStoreDB; + $compress = $storeObj->store( $extdb, $compress ); + if ( $compress === false ) { + print "Unable to store object\n"; + return false; + } + } + + # Update text row + $dbw->update( 'text', + array( /* SET */ + 'old_flags' => $flags, + 'old_text' => $compress + ), array( /* WHERE */ + 'old_id' => $row->old_id + ), $fname, 'LIMIT 1' + ); + return true; +} + +define( 'LS_INDIVIDUAL', 0 ); +define( 'LS_CHUNKED', 1 ); + +/** @todo document */ +function compressWithConcat( $startId, $maxChunkSize, $maxChunkFactor, $factorThreshold, $beginDate, + $endDate, $extdb="", $maxPageId = false ) +{ + $fname = 'compressWithConcat'; + $loadStyle = LS_CHUNKED; + + $dbr =& wfGetDB( DB_SLAVE ); + $dbw =& wfGetDB( DB_MASTER ); + + # Set up external storage + if ( $extdb != '' ) { + $storeObj = new ExternalStoreDB; + } + + # Get all articles by page_id + if ( !$maxPageId ) { + $maxPageId = $dbr->selectField( 'page', 'max(page_id)', '', $fname ); + } + print "Starting from $startId of $maxPageId\n"; + $pageConds = array(); + + /* + if ( $exclude_ns0 ) { + print "Excluding main namespace\n"; + $pageConds[] = 'page_namespace<>0'; + } + if ( $queryExtra ) { + $pageConds[] = $queryExtra; + } + */ + + # For each article, get a list of revisions which fit the criteria + + # No recompression, use a condition on old_flags + # Don't compress object type entities, because that might produce data loss when + # overwriting bulk storage concat rows. Don't compress external references, because + # the script doesn't yet delete rows from external storage. + $conds = array( + "old_flags NOT LIKE '%object%' AND old_flags NOT LIKE '%external%'"); + + if ( $beginDate ) { + $conds[] = "rev_timestamp>'" . $beginDate . "'"; + } + if ( $endDate ) { + $conds[] = "rev_timestamp<'" . $endDate . "'"; + } + if ( $loadStyle == LS_CHUNKED ) { + $tables = array( 'revision', 'text' ); + $fields = array( 'rev_id', 'rev_text_id', 'old_flags', 'old_text' ); + $conds[] = 'rev_text_id=old_id'; + $revLoadOptions = 'FOR UPDATE'; + } else { + $tables = array( 'revision' ); + $fields = array( 'rev_id', 'rev_text_id' ); + $revLoadOptions = array(); + } + + # Don't work with current revisions + # Don't lock the page table for update either -- TS 2006-04-04 + #$tables[] = 'page'; + #$conds[] = 'page_id=rev_page AND rev_id != page_latest'; + + $oldReadsSinceLastSlaveWait = 0; #check slave lag periodically + $totalMatchingRevisions = 0; + $masterPos = false; + for ( $pageId = $startId; $pageId <= $maxPageId; $pageId++ ) { + wfWaitForSlaves( 5 ); + + # Wake up + $dbr->ping(); + + # Get the page row + $pageRes = $dbr->select( 'page', + array('page_id', 'page_namespace', 'page_title','page_latest'), + $pageConds + array('page_id' => $pageId), $fname ); + if ( $dbr->numRows( $pageRes ) == 0 ) { + continue; + } + $pageRow = $dbr->fetchObject( $pageRes ); + + # Display progress + $titleObj = Title::makeTitle( $pageRow->page_namespace, $pageRow->page_title ); + print "$pageId\t" . $titleObj->getPrefixedDBkey() . " "; + + # Load revisions + $revRes = $dbw->select( $tables, $fields, + array( + 'rev_page' => $pageRow->page_id, + # Don't operate on the current revision + # Use < instead of <> in case the current revision has changed + # since the page select, which wasn't locking + 'rev_id < ' . $pageRow->page_latest + ) + $conds, + $fname, + $revLoadOptions + ); + $revs = array(); + while ( $revRow = $dbw->fetchObject( $revRes ) ) { + $revs[] = $revRow; + } + + if ( count( $revs ) < 2) { + # No revisions matching, no further processing + print "\n"; + continue; + } + + # For each chunk + $i = 0; + while ( $i < count( $revs ) ) { + if ( $i < count( $revs ) - $maxChunkSize ) { + $thisChunkSize = $maxChunkSize; + } else { + $thisChunkSize = count( $revs ) - $i; + } + + $chunk = new ConcatenatedGzipHistoryBlob(); + $stubs = array(); + $dbw->begin(); + $usedChunk = false; + $primaryOldid = $revs[$i]->rev_text_id; + + # Get the text of each revision and add it to the object + for ( $j = 0; $j < $thisChunkSize && $chunk->isHappy( $maxChunkFactor, $factorThreshold ); $j++ ) { + $oldid = $revs[$i + $j]->rev_text_id; + + # Get text + if ( $loadStyle == LS_INDIVIDUAL ) { + $textRow = $dbw->selectRow( 'text', + array( 'old_flags', 'old_text' ), + array( 'old_id' => $oldid ), + $fname, + 'FOR UPDATE' + ); + $text = Revision::getRevisionText( $textRow ); + } else { + $text = Revision::getRevisionText( $revs[$i + $j] ); + } + + if ( $text === false ) { + print "\nError, unable to get text in old_id $oldid\n"; + #$dbw->delete( 'old', array( 'old_id' => $oldid ) ); + } + + if ( $extdb == "" && $j == 0 ) { + $chunk->setText( $text ); + print '.'; + } else { + # Don't make a stub if it's going to be longer than the article + # Stubs are typically about 100 bytes + if ( strlen( $text ) < 120 ) { + $stub = false; + print 'x'; + } else { + $stub = $chunk->addItem( $text ); + $stub->setLocation( $primaryOldid ); + $stub->setReferrer( $oldid ); + print '.'; + $usedChunk = true; + } + $stubs[$j] = $stub; + } + } + $thisChunkSize = $j; + + # If we couldn't actually use any stubs because the pages were too small, do nothing + if ( $usedChunk ) { + if ( $extdb != "" ) { + # Move blob objects to External Storage + $stored = $storeObj->store( $extdb, serialize( $chunk )); + if ($stored === false) { + print "Unable to store object\n"; + return false; + } + # Store External Storage URLs instead of Stub placeholders + foreach ($stubs as $stub) { + if ($stub===false) + continue; + # $stored should provide base path to a BLOB + $url = $stored."/".$stub->getHash(); + $dbw->update( 'text', + array( /* SET */ + 'old_text' => $url, + 'old_flags' => 'external,utf-8', + ), array ( /* WHERE */ + 'old_id' => $stub->getReferrer(), + ) + ); + } + } else { + # Store the main object locally + $dbw->update( 'text', + array( /* SET */ + 'old_text' => serialize( $chunk ), + 'old_flags' => 'object,utf-8', + ), array( /* WHERE */ + 'old_id' => $primaryOldid + ) + ); + + # Store the stub objects + for ( $j = 1; $j < $thisChunkSize; $j++ ) { + # Skip if not compressing + if ( $stubs[$j] !== false ) { + $dbw->update( 'text', + array( /* SET */ + 'old_text' => serialize($stubs[$j]), + 'old_flags' => 'object,utf-8', + ), array( /* WHERE */ + 'old_id' => $revs[$i + $j]->rev_text_id + ) + ); + } + } + } + } + # Done, next + print "/"; + $dbw->commit(); + $i += $thisChunkSize; + wfWaitForSlaves( 5 ); + } + print "\n"; + } + return true; +} +?> -- cgit v1.2.3-54-g00ecf