maintenance/deduplicateArchiveRevId.php

   1 <?php
   2
   3 use Wikimedia\Rdbms\IDatabase;
   4
   5 require_once __DIR__ . '/Maintenance.php';
   6
   7 /**
   8  * Maintenance script that cleans up archive rows with duplicated ar_rev_id,
   9  * both within archive and between archive and revision.
  10  *
  11  * @ingroup Maintenance
  12  * @since 1.32
  13  */
  14 class DeduplicateArchiveRevId extends LoggedUpdateMaintenance {
  15
  16         /** @var array|null */
  17         private $arActorQuery = null;
  18
  19         private $deleted = 0;
  20         private $reassigned = 0;
  21
  22         public function __construct() {
  23                 parent::__construct();
  24                 $this->addDescription(
  25                         'Clean up duplicate ar_rev_id, both within archive and between archive and revision.'
  26                 );
  27                 $this->setBatchSize( 10000 );
  28         }
  29
  30         protected function getUpdateKey() {
  31                 return __CLASS__;
  32         }
  33
  34         protected function doDBUpdates() {
  35                 $this->output( "Deduplicating ar_rev_id...\n" );
  36
  37                 $dbw = $this->getDB( DB_MASTER );
  38
  39                 $minId = $dbw->selectField( 'archive', 'MIN(ar_rev_id)', [], __METHOD__ );
  40                 $maxId = $dbw->selectField( 'archive', 'MAX(ar_rev_id)', [], __METHOD__ );
  41                 $batchSize = $this->getBatchSize();
  42
  43                 $this->arActorQuery = ActorMigration::newMigration()->getJoin( 'ar_user' );
  44                 $revActorQuery = ActorMigration::newMigration()->getJoin( 'rev_user' );
  45
  46                 for ( $id = $minId; $id <= $maxId; $id += $batchSize ) {
  47                         $endId = min( $maxId, $id + $batchSize - 1 );
  48
  49                         $this->beginTransaction( $dbw, __METHOD__ );
  50
  51                         // Lock the archive and revision table rows for the IDs we're checking
  52                         // to try to prevent deletions or undeletions from confusing things.
  53                         $dbw->selectRowCount(
  54                                 'archive',
  55                                 1,
  56                                 [ 'ar_rev_id >= ' . (int)$id, 'ar_rev_id <= ' . (int)$endId ],
  57                                 __METHOD__,
  58                                 [ 'FOR UPDATE' ]
  59                         );
  60                         $dbw->selectRowCount(
  61                                 'revision',
  62                                 1,
  63                                 [ 'rev_id >= ' . (int)$id, 'rev_id <= ' . (int)$endId ],
  64                                 __METHOD__,
  65                                 [ 'LOCK IN SHARE MODE' ]
  66                         );
  67
  68                         // Figure out the ar_rev_ids we actually need to look at
  69                         $res = $dbw->select(
  70                                 [ 'archive', 'revision' ] + $revActorQuery['tables'],
  71                                 [ 'rev_id', 'rev_timestamp', 'rev_sha1' ] + $revActorQuery['fields'],
  72                                 [ 'ar_rev_id >= ' . (int)$id, 'ar_rev_id <= ' . (int)$endId ],
  73                                 __METHOD__,
  74                                 [ 'DISTINCT' ],
  75                                 [ 'revision' => [ 'JOIN', 'ar_rev_id = rev_id' ] ] + $revActorQuery['joins']
  76                         );
  77                         $revRows = [];
  78                         foreach ( $res as $row ) {
  79                                 $revRows[$row->rev_id] = $row;
  80                         }
  81
  82                         $arRevIds = $dbw->selectFieldValues(
  83                                 [ 'archive' ],
  84                                 'ar_rev_id',
  85                                 [ 'ar_rev_id >= ' . (int)$id, 'ar_rev_id <= ' . (int)$endId ],
  86                                 __METHOD__,
  87                                 [ 'GROUP BY' => 'ar_rev_id', 'HAVING' => 'COUNT(*) > 1' ]
  88                         );
  89                         $arRevIds = array_values( array_unique( array_merge( $arRevIds, array_keys( $revRows ) ) ) );
  90
  91                         if ( $arRevIds ) {
  92                                 $this->processArRevIds( $dbw, $arRevIds, $revRows );
  93                         }
  94
  95                         $this->output( "... $id-$endId\n" );
  96                         $this->commitTransaction( $dbw, __METHOD__ );
  97                 }
  98
  99                 $this->output(
 100                         "Finished deduplicating ar_rev_id. $this->deleted rows deleted, "
 101                         . "$this->reassigned assigned new IDs.\n"
 102                 );
 103                 return true;
 104         }
 105
 106         /**
 107          * Process a set of ar_rev_ids
 108          * @param IDatabase $dbw
 109          * @param int[] $arRevIds IDs to process
 110          * @param object[] $revRows Existing revision-table row data
 111          */
 112         private function processArRevIds( IDatabase $dbw, array $arRevIds, array $revRows ) {
 113                 // Select all the data we need for deduplication
 114                 $res = $dbw->select(
 115                         [ 'archive' ] + $this->arActorQuery['tables'],
 116                         [ 'ar_id', 'ar_rev_id', 'ar_namespace', 'ar_title', 'ar_timestamp', 'ar_sha1' ]
 117                                 + $this->arActorQuery['fields'],
 118                         [ 'ar_rev_id' => $arRevIds ],
 119                         __METHOD__,
 120                         [],
 121                         $this->arActorQuery['joins']
 122                 );
 123
 124                 // Determine which rows we need to delete or reassign
 125                 $seen = [];
 126                 $toDelete = [];
 127                 $toReassign = [];
 128                 foreach ( $res as $row ) {
 129                         // Revision-table row exists?
 130                         if ( isset( $revRows[$row->ar_rev_id] ) ) {
 131                                 $revRow = $revRows[$row->ar_rev_id];
 132
 133                                 // Record the rev_id as seen, so the code below will always delete or reassign.
 134                                 if ( !isset( $seen[$revRow->rev_id] ) ) {
 135                                         $seen[$revRow->rev_id] = [
 136                                                 'first' => "revision row",
 137                                         ];
 138                                 }
 139
 140                                 // Delete the archive row if it seems to be the same regardless
 141                                 // of page, because moves can change IDs and titles.
 142                                 if ( $row->ar_timestamp === $revRow->rev_timestamp &&
 143                                         $row->ar_sha1 === $revRow->rev_sha1 &&
 144                                         $row->ar_user === $revRow->rev_user &&
 145                                         $row->ar_user_text === $revRow->rev_user_text
 146                                 ) {
 147                                         $this->output(
 148                                                 "Row $row->ar_id duplicates revision row for rev_id $revRow->rev_id, deleting\n"
 149                                         );
 150                                         $toDelete[] = $row->ar_id;
 151                                         continue;
 152                                 }
 153                         }
 154
 155                         $key = $this->getSeenKey( $row );
 156                         if ( !isset( $seen[$row->ar_rev_id] ) ) {
 157                                 // This rev_id hasn't even been seen yet, nothing to do besides record it.
 158                                 $seen[$row->ar_rev_id] = [
 159                                         'first' => "archive row $row->ar_id",
 160                                         $key => $row->ar_id,
 161                                 ];
 162                         } elseif ( !isset( $seen[$row->ar_rev_id][$key] ) ) {
 163                                 // The rev_id was seen, but not this particular change. Reassign it.
 164                                 $seen[$row->ar_rev_id][$key] = $row->ar_id;
 165                                 $this->output(
 166                                         "Row $row->ar_id conflicts with {$seen[$row->ar_rev_id]['first']} "
 167                                         . "for rev_id $row->ar_rev_id, reassigning\n"
 168                                 );
 169                                 $toReassign[] = $row->ar_id;
 170                         } else {
 171                                 // The rev_id was seen with a row that matches this change. Delete it.
 172                                 $this->output(
 173                                         "Row $row->ar_id duplicates archive row {$seen[$row->ar_rev_id][$key]} "
 174                                         . "for rev_id $row->ar_rev_id, deleting\n"
 175                                 );
 176                                 $toDelete[] = $row->ar_id;
 177                         }
 178                 }
 179
 180                 // Perform the updates
 181                 if ( $toDelete ) {
 182                         $dbw->delete( 'archive', [ 'ar_id' => $toDelete ], __METHOD__ );
 183                         $this->deleted += $dbw->affectedRows();
 184                 }
 185                 if ( $toReassign ) {
 186                         $this->reassigned += PopulateArchiveRevId::reassignArRevIds( $dbw, $toReassign );
 187                 }
 188         }
 189
 190         /**
 191          * Make a key identifying a "unique" change from a row
 192          * @param object $row
 193          * @return string
 194          */
 195         private function getSeenKey( $row ) {
 196                 return implode( "\n", [
 197                         $row->ar_namespace,
 198                         $row->ar_title,
 199                         $row->ar_timestamp,
 200                         $row->ar_sha1,
 201                         $row->ar_user,
 202                         $row->ar_user_text,
 203                 ] );
 204         }
 205
 206 }
 207
 208 $maintClass = "DeduplicateArchiveRevId";
 209 require_once RUN_MAINTENANCE_IF_MAIN;