maintenance/cleanupImages.php

   1 <?php
   2 /**
   3  * Clean up broken, unparseable upload filenames.
   4  *
   5  * Copyright © 2005-2006 Brion Vibber <brion@pobox.com>
   6  * https://www.mediawiki.org/
   7  *
   8  * This program is free software; you can redistribute it and/or modify
   9  * it under the terms of the GNU General Public License as published by
  10  * the Free Software Foundation; either version 2 of the License, or
  11  * (at your option) any later version.
  12  *
  13  * This program is distributed in the hope that it will be useful,
  14  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  15  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  16  * GNU General Public License for more details.
  17  *
  18  * You should have received a copy of the GNU General Public License along
  19  * with this program; if not, write to the Free Software Foundation, Inc.,
  20  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  21  * http://www.gnu.org/copyleft/gpl.html
  22  *
  23  * @file
  24  * @author Brion Vibber <brion at pobox.com>
  25  * @ingroup Maintenance
  26  */
  27
  28 require_once __DIR__ . '/cleanupTable.inc';
  29
  30 /**
  31  * Maintenance script to clean up broken, unparseable upload filenames.
  32  *
  33  * @ingroup Maintenance
  34  */
  35 class ImageCleanup extends TableCleanup {
  36         protected $defaultParams = [
  37                 'table' => 'image',
  38                 'conds' => [],
  39                 'index' => 'img_name',
  40                 'callback' => 'processRow',
  41         ];
  42
  43         public function __construct() {
  44                 parent::__construct();
  45                 $this->addDescription( 'Script to clean up broken, unparseable upload filenames' );
  46         }
  47
  48         protected function processRow( $row ) {
  49                 global $wgContLang;
  50
  51                 $source = $row->img_name;
  52                 if ( $source == '' ) {
  53                         // Ye olde empty rows. Just kill them.
  54                         $this->killRow( $source );
  55
  56                         return $this->progress( 1 );
  57                 }
  58
  59                 $cleaned = $source;
  60
  61                 // About half of old bad image names have percent-codes
  62                 $cleaned = rawurldecode( $cleaned );
  63
  64                 // We also have some HTML entities there
  65                 $cleaned = Sanitizer::decodeCharReferences( $cleaned );
  66
  67                 // Some are old latin-1
  68                 $cleaned = $wgContLang->checkTitleEncoding( $cleaned );
  69
  70                 // Many of remainder look like non-normalized unicode
  71                 $cleaned = $wgContLang->normalize( $cleaned );
  72
  73                 $title = Title::makeTitleSafe( NS_FILE, $cleaned );
  74
  75                 if ( is_null( $title ) ) {
  76                         $this->output( "page $source ($cleaned) is illegal.\n" );
  77                         $safe = $this->buildSafeTitle( $cleaned );
  78                         if ( $safe === false ) {
  79                                 return $this->progress( 0 );
  80                         }
  81                         $this->pokeFile( $source, $safe );
  82
  83                         return $this->progress( 1 );
  84                 }
  85
  86                 if ( $title->getDBkey() !== $source ) {
  87                         $munged = $title->getDBkey();
  88                         $this->output( "page $source ($munged) doesn't match self.\n" );
  89                         $this->pokeFile( $source, $munged );
  90
  91                         return $this->progress( 1 );
  92                 }
  93
  94                 return $this->progress( 0 );
  95         }
  96
  97         /**
  98          * @param string $name
  99          */
 100         private function killRow( $name ) {
 101                 if ( $this->dryrun ) {
 102                         $this->output( "DRY RUN: would delete bogus row '$name'\n" );
 103                 } else {
 104                         $this->output( "deleting bogus row '$name'\n" );
 105                         $db = $this->getDB( DB_MASTER );
 106                         $db->delete( 'image',
 107                                 [ 'img_name' => $name ],
 108                                 __METHOD__ );
 109                 }
 110         }
 111
 112         private function filePath( $name ) {
 113                 if ( !isset( $this->repo ) ) {
 114                         $this->repo = RepoGroup::singleton()->getLocalRepo();
 115                 }
 116
 117                 return $this->repo->getRootDirectory() . '/' . $this->repo->getHashPath( $name ) . $name;
 118         }
 119
 120         private function imageExists( $name, $db ) {
 121                 return $db->selectField( 'image', '1', [ 'img_name' => $name ], __METHOD__ );
 122         }
 123
 124         private function pageExists( $name, $db ) {
 125                 return $db->selectField(
 126                         'page',
 127                         '1',
 128                         [ 'page_namespace' => NS_FILE, 'page_title' => $name ],
 129                         __METHOD__
 130                 );
 131         }
 132
 133         private function pokeFile( $orig, $new ) {
 134                 $path = $this->filePath( $orig );
 135                 if ( !file_exists( $path ) ) {
 136                         $this->output( "missing file: $path\n" );
 137                         $this->killRow( $orig );
 138
 139                         return;
 140                 }
 141
 142                 $db = $this->getDB( DB_MASTER );
 143
 144                 /*
 145                  * To prevent key collisions in the update() statements below,
 146                  * if the target title exists in the image table, or if both the
 147                  * original and target titles exist in the page table, append
 148                  * increasing version numbers until the target title exists in
 149                  * neither.  (See also T18916.)
 150                  */
 151                 $version = 0;
 152                 $final = $new;
 153                 $conflict = ( $this->imageExists( $final, $db ) ||
 154                         ( $this->pageExists( $orig, $db ) && $this->pageExists( $final, $db ) ) );
 155
 156                 while ( $conflict ) {
 157                         $this->output( "Rename conflicts with '$final'...\n" );
 158                         $version++;
 159                         $final = $this->appendTitle( $new, "_$version" );
 160                         $conflict = ( $this->imageExists( $final, $db ) || $this->pageExists( $final, $db ) );
 161                 }
 162
 163                 $finalPath = $this->filePath( $final );
 164
 165                 if ( $this->dryrun ) {
 166                         $this->output( "DRY RUN: would rename $path to $finalPath\n" );
 167                 } else {
 168                         $this->output( "renaming $path to $finalPath\n" );
 169                         // @todo FIXME: Should this use File::move()?
 170                         $this->beginTransaction( $db, __METHOD__ );
 171                         $db->update( 'image',
 172                                 [ 'img_name' => $final ],
 173                                 [ 'img_name' => $orig ],
 174                                 __METHOD__ );
 175                         $db->update( 'oldimage',
 176                                 [ 'oi_name' => $final ],
 177                                 [ 'oi_name' => $orig ],
 178                                 __METHOD__ );
 179                         $db->update( 'page',
 180                                 [ 'page_title' => $final ],
 181                                 [ 'page_title' => $orig, 'page_namespace' => NS_FILE ],
 182                                 __METHOD__ );
 183                         $dir = dirname( $finalPath );
 184                         if ( !file_exists( $dir ) ) {
 185                                 if ( !wfMkdirParents( $dir, null, __METHOD__ ) ) {
 186                                         $this->output( "RENAME FAILED, COULD NOT CREATE $dir" );
 187                                         $this->rollbackTransaction( $db, __METHOD__ );
 188
 189                                         return;
 190                                 }
 191                         }
 192                         if ( rename( $path, $finalPath ) ) {
 193                                 $this->commitTransaction( $db, __METHOD__ );
 194                         } else {
 195                                 $this->error( "RENAME FAILED" );
 196                                 $this->rollbackTransaction( $db, __METHOD__ );
 197                         }
 198                 }
 199         }
 200
 201         private function appendTitle( $name, $suffix ) {
 202                 return preg_replace( '/^(.*)(\..*?)$/',
 203                         "\\1$suffix\\2", $name );
 204         }
 205
 206         private function buildSafeTitle( $name ) {
 207                 $x = preg_replace_callback(
 208                         '/([^' . Title::legalChars() . ']|~)/',
 209                         [ $this, 'hexChar' ],
 210                         $name );
 211
 212                 $test = Title::makeTitleSafe( NS_FILE, $x );
 213                 if ( is_null( $test ) || $test->getDBkey() !== $x ) {
 214                         $this->error( "Unable to generate safe title from '$name', got '$x'" );
 215
 216                         return false;
 217                 }
 218
 219                 return $x;
 220         }
 221 }
 222
 223 $maintClass = "ImageCleanup";
 224 require_once RUN_MAINTENANCE_IF_MAIN;