maintenance/cleanupImages.php

   1 <?php
   2 /**
   3  * Script to clean up broken, unparseable upload filenames.
   4  *
   5  * Usage: php cleanupImages.php [--fix]
   6  * Options:
   7  *   --fix  Actually clean up titles; otherwise just checks for them
   8  *
   9  * Copyright (C) 2005-2006 Brion Vibber <brion@pobox.com>
  10  * http://www.mediawiki.org/
  11  *
  12  * This program is free software; you can redistribute it and/or modify
  13  * it under the terms of the GNU General Public License as published by
  14  * the Free Software Foundation; either version 2 of the License, or
  15  * (at your option) any later version.
  16  *
  17  * This program is distributed in the hope that it will be useful,
  18  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  19  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  20  * GNU General Public License for more details.
  21  *
  22  * You should have received a copy of the GNU General Public License along
  23  * with this program; if not, write to the Free Software Foundation, Inc.,
  24  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  25  * http://www.gnu.org/copyleft/gpl.html
  26  *
  27  * @file
  28  * @author Brion Vibber <brion at pobox.com>
  29  * @ingroup Maintenance
  30  */
  31
  32 require_once( dirname( __FILE__ ) . '/cleanupTable.inc' );
  33
  34 class ImageCleanup extends TableCleanup {
  35         protected $defaultParams = array(
  36                 'table' => 'image',
  37                 'conds' => array(),
  38                 'index' => 'img_name',
  39                 'callback' => 'processRow',
  40         );
  41
  42         public function __construct() {
  43                 parent::__construct();
  44                 $this->mDescription = "Script to clean up broken, unparseable upload filenames";
  45         }
  46
  47         protected function processRow( $row ) {
  48                 global $wgContLang;
  49
  50                 $source = $row->img_name;
  51                 if ( $source == '' ) {
  52                         // Ye olde empty rows. Just kill them.
  53                         $this->killRow( $source );
  54                         return $this->progress( 1 );
  55                 }
  56
  57                 $cleaned = $source;
  58
  59                 // About half of old bad image names have percent-codes
  60                 $cleaned = rawurldecode( $cleaned );
  61
  62                 // We also have some HTML entities there
  63                 $cleaned = Sanitizer::decodeCharReferences( $cleaned );
  64
  65                 // Some are old latin-1
  66                 $cleaned = $wgContLang->checkTitleEncoding( $cleaned );
  67
  68                 // Many of remainder look like non-normalized unicode
  69                 $cleaned = $wgContLang->normalize( $cleaned );
  70
  71                 $title = Title::makeTitleSafe( NS_FILE, $cleaned );
  72
  73                 if ( is_null( $title ) ) {
  74                         $this->output( "page $source ($cleaned) is illegal.\n" );
  75                         $safe = $this->buildSafeTitle( $cleaned );
  76                         if ( $safe === false )
  77                                 return $this->progress( 0 );
  78                         $this->pokeFile( $source, $safe );
  79                         return $this->progress( 1 );
  80                 }
  81
  82                 if ( $title->getDBkey() !== $source ) {
  83                         $munged = $title->getDBkey();
  84                         $this->output( "page $source ($munged) doesn't match self.\n" );
  85                         $this->pokeFile( $source, $munged );
  86                         return $this->progress( 1 );
  87                 }
  88
  89                 $this->progress( 0 );
  90         }
  91
  92         private function killRow( $name ) {
  93                 if ( $this->dryrun ) {
  94                         $this->output( "DRY RUN: would delete bogus row '$name'\n" );
  95                 } else {
  96                         $this->output( "deleting bogus row '$name'\n" );
  97                         $db = wfGetDB( DB_MASTER );
  98                         $db->delete( 'image',
  99                                 array( 'img_name' => $name ),
 100                                 __METHOD__ );
 101                 }
 102         }
 103
 104         private function filePath( $name ) {
 105                 if ( !isset( $this->repo ) ) {
 106                         $this->repo = RepoGroup::singleton()->getLocalRepo();
 107                 }
 108                 return $this->repo->getRootDirectory() . '/' . $this->repo->getHashPath( $name ) . $name;
 109         }
 110
 111         private function imageExists( $name, $db ) {
 112                 return $db->selectField( 'image', '1', array( 'img_name' => $name ), __METHOD__ );
 113         }
 114
 115         private function pageExists( $name, $db ) {
 116                 return $db->selectField( 'page', '1', array( 'page_namespace' => NS_FILE, 'page_title' => $name ), __METHOD__ );
 117         }
 118
 119         private function pokeFile( $orig, $new ) {
 120                 $path = $this->filePath( $orig );
 121                 if ( !file_exists( $path ) ) {
 122                         $this->output( "missing file: $path\n" );
 123                         return $this->killRow( $orig );
 124                 }
 125
 126                 $db = wfGetDB( DB_MASTER );
 127
 128                 /*
 129                  * To prevent key collisions in the update() statements below,
 130                  * if the target title exists in the image table, or if both the
 131                  * original and target titles exist in the page table, append
 132                  * increasing version numbers until the target title exists in
 133                  * neither.  (See also bug 16916.)
 134                  */
 135                 $version = 0;
 136                 $final = $new;
 137                 $conflict = ( $this->imageExists( $final, $db ) ||
 138                                   ( $this->pageExists( $orig, $db ) && $this->pageExists( $final, $db ) ) );
 139
 140                 while ( $conflict ) {
 141                         $this->output( "Rename conflicts with '$final'...\n" );
 142                         $version++;
 143                         $final = $this->appendTitle( $new, "_$version" );
 144                         $conflict = ( $this->imageExists( $final, $db ) || $this->pageExists( $final, $db ) );
 145                 }
 146
 147                 $finalPath = $this->filePath( $final );
 148
 149                 if ( $this->dryrun ) {
 150                         $this->output( "DRY RUN: would rename $path to $finalPath\n" );
 151                 } else {
 152                         $this->output( "renaming $path to $finalPath\n" );
 153                         // XXX: should this use File::move()?  FIXME?
 154                         $db->begin();
 155                         $db->update( 'image',
 156                                 array( 'img_name' => $final ),
 157                                 array( 'img_name' => $orig ),
 158                                 __METHOD__ );
 159                         $db->update( 'oldimage',
 160                                 array( 'oi_name' => $final ),
 161                                 array( 'oi_name' => $orig ),
 162                                 __METHOD__ );
 163                         $db->update( 'page',
 164                                 array( 'page_title' => $final ),
 165                                 array( 'page_title' => $orig, 'page_namespace' => NS_FILE ),
 166                                 __METHOD__ );
 167                         $dir = dirname( $finalPath );
 168                         if ( !file_exists( $dir ) ) {
 169                                 if ( !wfMkdirParents( $dir ) ) {
 170                                         $this->log( "RENAME FAILED, COULD NOT CREATE $dir" );
 171                                         $db->rollback();
 172                                         return;
 173                                 }
 174                         }
 175                         if ( rename( $path, $finalPath ) ) {
 176                                 $db->commit();
 177                         } else {
 178                                 $this->error( "RENAME FAILED" );
 179                                 $db->rollback();
 180                         }
 181                 }
 182         }
 183
 184         private function appendTitle( $name, $suffix ) {
 185                 return preg_replace( '/^(.*)(\..*?)$/',
 186                         "\\1$suffix\\2", $name );
 187         }
 188
 189         private function buildSafeTitle( $name ) {
 190                 global $wgLegalTitleChars;
 191                 $x = preg_replace_callback(
 192                         "/([^$wgLegalTitleChars]|~)/",
 193                         array( $this, 'hexChar' ),
 194                         $name );
 195
 196                 $test = Title::makeTitleSafe( NS_FILE, $x );
 197                 if ( is_null( $test ) || $test->getDBkey() !== $x ) {
 198                         $this->error( "Unable to generate safe title from '$name', got '$x'" );
 199                         return false;
 200                 }
 201
 202                 return $x;
 203         }
 204 }
 205
 206 $maintClass = "ImageCleanup";
 207 require_once( RUN_MAINTENANCE_IF_MAIN );