From: Tim Starling Date: Sun, 14 Aug 2005 07:57:05 +0000 (+0000) Subject: Moving compression scripts from maintenance to maintenance/storage X-Git-Tag: 1.6.0~1987 X-Git-Url: https://git.cyclocoop.org/%7B%24www_url%7Dadmin/compta/exercices/journal.php?a=commitdiff_plain;h=31c064f436c8c551d921fa238d3172b60ae0faba;p=lhc%2Fweb%2Fwiklou.git Moving compression scripts from maintenance to maintenance/storage --- diff --git a/maintenance/compressOld.inc b/maintenance/compressOld.inc deleted file mode 100644 index f6a9f43035..0000000000 --- a/maintenance/compressOld.inc +++ /dev/null @@ -1,230 +0,0 @@ -tableName( 'old' ); - do { - $end = $start + $chunksize; - $res = $dbw->select( 'old', array( 'old_id','old_flags','old_namespace','old_title','old_text' ), - "old_id>=$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 ); - $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 ) { - $fname = 'compressPage'; - if( false !== strpos( $row->old_flags, "gzip" ) ) { - print "Already compressed row {$row->old_id}?\n"; - return false; - } - $dbw =& wfGetDB( DB_MASTER ); - $flags = $row->old_flags ? "{$row->old_flags},gzip" : "gzip"; - $compress = gzdeflate( $row->old_text ); - $dbw->update( 'old', - 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 ) -{ - $fname = 'compressWithConcat'; - $loadStyle = LS_CHUNKED; - - $dbr =& wfGetDB( DB_SLAVE ); - $dbw =& wfGetDB( DB_MASTER ); - - # Get all articles by page_id - $maxPageId = $dbr->selectField( 'page', 'max(page_id)', '', $fname ); - $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 - $conds = array("old_flags NOT LIKE '%object%'"); - - 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(); - } - - $oldReadsSinceLastSlaveWait = 0; #check slave lag periodically - $totalMatchingRevisions = 0; - $masterPos = false; - for ( $pageId = $startId; $pageId <= $maxPageId; $pageId++ ) { - $pageRes = $dbr->select( 'page', array('page_id', 'page_namespace', 'page_title'), - $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 ) + $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 ( $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 ); - $hash = $stub->getHash(); - $stub = serialize( $stub ); - 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 ) { - # Store the main object - $dbw->update( 'text', - array( /* SET */ - 'old_text' => serialize( $chunk ), - 'old_flags' => 'object', - ), 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' => $stubs[$j], - 'old_flags' => 'object', - ), array( /* WHERE */ - 'old_id' => $revs[$i + $j]->rev_text_id - ) - ); - } - } - } - # Done, next - print "/"; - $dbw->commit(); - $i += $thisChunkSize; - } - print "\n"; - } - return true; -} -?> diff --git a/maintenance/compressOld.php b/maintenance/compressOld.php deleted file mode 100644 index 0d449c4c7c..0000000000 --- a/maintenance/compressOld.php +++ /dev/null @@ -1,83 +0,0 @@ -] [-c ] [-b ] [-e ] [-s ] - * - * Wikimedia - * php compressOld.php [-t ] [-c ] [-b ] [-e ] [-s ] - * [-f ] [-h ] - * - * is either: - * gzip: compress revisions independently - * concat: concatenate revisions and compress in chunks (default) - * - * is the old_id to start from - * - * The following options apply only to the concat type: - * is the earliest date to check for uncompressed revisions - * is the latest revision date to compress - * is the maximum number of revisions in a concat chunk - * is the maximum ratio of compressed chunk bytes to uncompressed avg. revision bytes - * is a minimum number of KB, where cuts in - * - */ - -die( 'compressOld is known to be broken at the moment.' ); - -$optionsWithArgs = array( 't', 'c', 's', 'f', 'h' ); -require_once( "commandLine.inc" ); -require_once( "compressOld.inc" ); - -if( !function_exists( "gzdeflate" ) ) { - print "You must enable zlib support in PHP to compress old revisions!\n"; - print "Please see http://www.php.net/manual/en/ref.zlib.php\n\n"; - die(); -} - -$defaults = array( - 't' => 'concat', - 'c' => 20, - 's' => 0, - 'f' => 3, - 'h' => 100, - 'b' => '', - 'e' => '', -); - -$args = $args + $defaults; - -if ( $args['t'] != 'concat' && $args['t'] != 'gzip' ) { - print "Type \"{$args['t']}\" not supported\n"; -} - -print "Depending on the size of your database this may take a while!\n"; -print "If you abort the script while it's running it shouldn't harm anything,\n"; -print "but if you haven't backed up your data, you SHOULD abort now!\n\n"; -print "Press control-c to abort first (will proceed automatically in 5 seconds)\n"; -#sleep(5); - -$success = true; -if ( $args['t'] == 'concat' ) { - $success = compressWithConcat( $args['s'], $args['c'], $args['f'], $args['h'], $args['b'], $args['e'] ); -} else { - compressOldPages( $args['s'] ); -} - -if ( $success ) { - print "Done.\n"; -} - -exit(); - -?> diff --git a/maintenance/storage/compressOld.inc b/maintenance/storage/compressOld.inc new file mode 100644 index 0000000000..f6a9f43035 --- /dev/null +++ b/maintenance/storage/compressOld.inc @@ -0,0 +1,230 @@ +tableName( 'old' ); + do { + $end = $start + $chunksize; + $res = $dbw->select( 'old', array( 'old_id','old_flags','old_namespace','old_title','old_text' ), + "old_id>=$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 ); + $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 ) { + $fname = 'compressPage'; + if( false !== strpos( $row->old_flags, "gzip" ) ) { + print "Already compressed row {$row->old_id}?\n"; + return false; + } + $dbw =& wfGetDB( DB_MASTER ); + $flags = $row->old_flags ? "{$row->old_flags},gzip" : "gzip"; + $compress = gzdeflate( $row->old_text ); + $dbw->update( 'old', + 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 ) +{ + $fname = 'compressWithConcat'; + $loadStyle = LS_CHUNKED; + + $dbr =& wfGetDB( DB_SLAVE ); + $dbw =& wfGetDB( DB_MASTER ); + + # Get all articles by page_id + $maxPageId = $dbr->selectField( 'page', 'max(page_id)', '', $fname ); + $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 + $conds = array("old_flags NOT LIKE '%object%'"); + + 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(); + } + + $oldReadsSinceLastSlaveWait = 0; #check slave lag periodically + $totalMatchingRevisions = 0; + $masterPos = false; + for ( $pageId = $startId; $pageId <= $maxPageId; $pageId++ ) { + $pageRes = $dbr->select( 'page', array('page_id', 'page_namespace', 'page_title'), + $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 ) + $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 ( $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 ); + $hash = $stub->getHash(); + $stub = serialize( $stub ); + 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 ) { + # Store the main object + $dbw->update( 'text', + array( /* SET */ + 'old_text' => serialize( $chunk ), + 'old_flags' => 'object', + ), 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' => $stubs[$j], + 'old_flags' => 'object', + ), array( /* WHERE */ + 'old_id' => $revs[$i + $j]->rev_text_id + ) + ); + } + } + } + # Done, next + print "/"; + $dbw->commit(); + $i += $thisChunkSize; + } + print "\n"; + } + return true; +} +?> diff --git a/maintenance/storage/compressOld.php b/maintenance/storage/compressOld.php new file mode 100644 index 0000000000..cde284b788 --- /dev/null +++ b/maintenance/storage/compressOld.php @@ -0,0 +1,83 @@ +] [-c ] [-b ] [-e ] [-s ] + * + * Wikimedia + * php compressOld.php [-t ] [-c ] [-b ] [-e ] [-s ] + * [-f ] [-h ] + * + * is either: + * gzip: compress revisions independently + * concat: concatenate revisions and compress in chunks (default) + * + * is the old_id to start from + * + * The following options apply only to the concat type: + * is the earliest date to check for uncompressed revisions + * is the latest revision date to compress + * is the maximum number of revisions in a concat chunk + * is the maximum ratio of compressed chunk bytes to uncompressed avg. revision bytes + * is a minimum number of KB, where cuts in + * + */ + +die( 'compressOld is known to be broken at the moment.' ); + +$optionsWithArgs = array( 't', 'c', 's', 'f', 'h' ); +require_once( "../commandLine.inc" ); +require_once( "compressOld.inc" ); + +if( !function_exists( "gzdeflate" ) ) { + print "You must enable zlib support in PHP to compress old revisions!\n"; + print "Please see http://www.php.net/manual/en/ref.zlib.php\n\n"; + die(); +} + +$defaults = array( + 't' => 'concat', + 'c' => 20, + 's' => 0, + 'f' => 3, + 'h' => 100, + 'b' => '', + 'e' => '', +); + +$args = $args + $defaults; + +if ( $args['t'] != 'concat' && $args['t'] != 'gzip' ) { + print "Type \"{$args['t']}\" not supported\n"; +} + +print "Depending on the size of your database this may take a while!\n"; +print "If you abort the script while it's running it shouldn't harm anything,\n"; +print "but if you haven't backed up your data, you SHOULD abort now!\n\n"; +print "Press control-c to abort first (will proceed automatically in 5 seconds)\n"; +#sleep(5); + +$success = true; +if ( $args['t'] == 'concat' ) { + $success = compressWithConcat( $args['s'], $args['c'], $args['f'], $args['h'], $args['b'], $args['e'] ); +} else { + compressOldPages( $args['s'] ); +} + +if ( $success ) { + print "Done.\n"; +} + +exit(); + +?>