From ce8925a1a935ee827a7d1221edcf39fff7979896 Mon Sep 17 00:00:00 2001 From: Brion Vibber Date: Thu, 1 May 2008 22:38:22 +0000 Subject: [PATCH] * (bug 13820) Fix updater for rev_parent_id population Was running a command-line entry point from the middle of an updater, which is bad bad bad -- will kill the web installer and may fail on command-line updates, depending on circumstances. --- RELEASE-NOTES | 2 + maintenance/populateParentId.inc | 81 +++++++++++++++++++++++++++++++ maintenance/populateParentId.php | 82 +------------------------------- maintenance/updaters.inc | 3 ++ 4 files changed, 87 insertions(+), 81 deletions(-) create mode 100644 maintenance/populateParentId.inc diff --git a/RELEASE-NOTES b/RELEASE-NOTES index 6e94c9e27f..9dc679af8c 100644 --- a/RELEASE-NOTES +++ b/RELEASE-NOTES @@ -236,6 +236,8 @@ it from source control: http://www.mediawiki.org/wiki/Download_from_SVN * (bug 13905) Blacklist Mac IE from HttpOnly cookies; it eats them sometimes * (bug 13922) Fix bad HTML on empty Special:Prefixindex and Special:Allpages * (bug 13924) Fix bad HTML on power search form +* (bug 13820) Fix updater for rev_parent_id population + === API changes in 1.13 === diff --git a/maintenance/populateParentId.inc b/maintenance/populateParentId.inc new file mode 100644 index 0000000000..c71d3119c8 --- /dev/null +++ b/maintenance/populateParentId.inc @@ -0,0 +1,81 @@ +selectField( 'revision', 'MIN(rev_id)', false, __FUNCTION__ ); + $end = $db->selectField( 'revision', 'MAX(rev_id)', false, __FUNCTION__ ); + if( is_null( $start ) || is_null( $end ) ){ + echo "...revision table seems to be empty.\n"; + $db->insert( 'updatelog', + array( 'ul_key' => 'populate rev_parent_id' ), + __FUNCTION__, + 'IGNORE' ); + return; + } + # Do remaining chunk + $end += BATCH_SIZE - 1; + $blockStart = $start; + $blockEnd = $start + BATCH_SIZE - 1; + $count = 0; + $changed = 0; + while( $blockEnd <= $end ) { + echo "...doing rev_id from $blockStart to $blockEnd\n"; + $cond = "rev_id BETWEEN $blockStart AND $blockEnd"; + $res = $db->select( 'revision', + array('rev_id','rev_page','rev_timestamp','rev_parent_id'), + $cond, __FUNCTION__ ); + # Go through and update rev_parent_id from these rows. + # Assume that the previous revision of the title was + # the original previous revision of the title when the + # edit was made... + foreach( $res as $row ) { + # First, check rows with the same timestamp other than this one + # with a smaller rev ID. The highest ID "wins". This avoids loops + # as timestamp can only decrease and never loops with IDs (from parent to parent) + $previousID = $db->selectField( 'revision', 'rev_id', + array( 'rev_page' => $row->rev_page, 'rev_timestamp' => $row->rev_timestamp, + "rev_id < {$row->rev_id}" ), + __FUNCTION__, + array( 'ORDER BY' => 'rev_id DESC' ) ); + # If there are none, check the the highest ID with a lower timestamp + if( !$previousID ) { + # Get the highest older timestamp + $lastTimestamp = $db->selectField( 'revision', 'rev_timestamp', + array( 'rev_page' => $row->rev_page, "rev_timestamp < '{$row->rev_timestamp}'" ), + __FUNCTION__, + array( 'ORDER BY' => 'rev_timestamp DESC' ) ); + # If there is one, let the highest rev ID win + if( $lastTimestamp ) { + $previousID = $db->selectField( 'revision', 'rev_id', + array( 'rev_page' => $row->rev_page, 'rev_timestamp' => $lastTimestamp ), + __FUNCTION__, + array( 'ORDER BY' => 'rev_id DESC' ) ); + } + } + $previousID = intval($previousID); + if( $previousID != $row->rev_parent_id ) + $changed++; + # Update the row... + $db->update( 'revision', + array( 'rev_parent_id' => $previousID ), + array( 'rev_id' => $row->rev_id ), + __FUNCTION__ ); + $count++; + } + $blockStart += BATCH_SIZE - 1; + $blockEnd += BATCH_SIZE - 1; + wfWaitForSlaves( 5 ); + } + $logged = $db->insert( 'updatelog', + array( 'ul_key' => 'populate rev_parent_id' ), + __FUNCTION__, + 'IGNORE' ); + if( $logged ) { + echo "rev_parent_id population complete ... {$count} rows [{$changed} changed]\n"; + return true; + } else { + echo "Could not insert rev_parent_id population row.\n"; + return false; + } +} + diff --git a/maintenance/populateParentId.php b/maintenance/populateParentId.php index 366c457c2b..b9d84750d7 100644 --- a/maintenance/populateParentId.php +++ b/maintenance/populateParentId.php @@ -9,6 +9,7 @@ define( 'BATCH_SIZE', 200 ); require_once 'commandLine.inc'; +require_once 'populateParentId.inc'; $db =& wfGetDB( DB_MASTER ); if ( !$db->tableExists( 'revision' ) ) { @@ -17,84 +18,3 @@ if ( !$db->tableExists( 'revision' ) ) { } populate_rev_parent_id( $db ); - -function populate_rev_parent_id( $db ) { - echo "Populating rev_parent_id column\n"; - $start = $db->selectField( 'revision', 'MIN(rev_id)', false, __FUNCTION__ ); - $end = $db->selectField( 'revision', 'MAX(rev_id)', false, __FUNCTION__ ); - if( is_null( $start ) || is_null( $end ) ){ - echo "...revision table seems to be empty.\n"; - $db->insert( 'updatelog', - array( 'ul_key' => 'populate rev_parent_id' ), - __FUNCTION__, - 'IGNORE' ); - return; - } - # Do remaining chunk - $end += BATCH_SIZE - 1; - $blockStart = $start; - $blockEnd = $start + BATCH_SIZE - 1; - $count = 0; - $changed = 0; - while( $blockEnd <= $end ) { - echo "...doing rev_id from $blockStart to $blockEnd\n"; - $cond = "rev_id BETWEEN $blockStart AND $blockEnd"; - $res = $db->select( 'revision', - array('rev_id','rev_page','rev_timestamp','rev_parent_id'), - $cond, __FUNCTION__ ); - # Go through and update rev_parent_id from these rows. - # Assume that the previous revision of the title was - # the original previous revision of the title when the - # edit was made... - foreach( $res as $row ) { - # First, check rows with the same timestamp other than this one - # with a smaller rev ID. The highest ID "wins". This avoids loops - # as timestamp can only decrease and never loops with IDs (from parent to parent) - $previousID = $db->selectField( 'revision', 'rev_id', - array( 'rev_page' => $row->rev_page, 'rev_timestamp' => $row->rev_timestamp, - "rev_id < {$row->rev_id}" ), - __FUNCTION__, - array( 'ORDER BY' => 'rev_id DESC' ) ); - # If there are none, check the the highest ID with a lower timestamp - if( !$previousID ) { - # Get the highest older timestamp - $lastTimestamp = $db->selectField( 'revision', 'rev_timestamp', - array( 'rev_page' => $row->rev_page, "rev_timestamp < '{$row->rev_timestamp}'" ), - __FUNCTION__, - array( 'ORDER BY' => 'rev_timestamp DESC' ) ); - # If there is one, let the highest rev ID win - if( $lastTimestamp ) { - $previousID = $db->selectField( 'revision', 'rev_id', - array( 'rev_page' => $row->rev_page, 'rev_timestamp' => $lastTimestamp ), - __FUNCTION__, - array( 'ORDER BY' => 'rev_id DESC' ) ); - } - } - $previousID = intval($previousID); - if( $previousID != $row->rev_parent_id ) - $changed++; - # Update the row... - $db->update( 'revision', - array( 'rev_parent_id' => $previousID ), - array( 'rev_id' => $row->rev_id ), - __FUNCTION__ ); - $count++; - } - $blockStart += BATCH_SIZE - 1; - $blockEnd += BATCH_SIZE - 1; - wfWaitForSlaves( 5 ); - } - $logged = $db->insert( 'updatelog', - array( 'ul_key' => 'populate rev_parent_id' ), - __FUNCTION__, - 'IGNORE' ); - if( $logged ) { - echo "rev_parent_id population complete ... {$count} rows [{$changed} changed]\n"; - return true; - } else { - echo "Could not insert rev_parent_id population row.\n"; - return false; - } -} - - diff --git a/maintenance/updaters.inc b/maintenance/updaters.inc index d2846636ea..86db76ffc0 100644 --- a/maintenance/updaters.inc +++ b/maintenance/updaters.inc @@ -1177,6 +1177,9 @@ function do_populate_parent_id() { return; } require_once( 'populateParentId.php' ); + + global $wgDatabase; + populate_rev_parent_id( $wgDatabase ); } function -- 2.20.1