maintenance/refreshLinks.php

   1 <?php
   2 /**
   3  * Refresh link tables.
   4  *
   5  * This program is free software; you can redistribute it and/or modify
   6  * it under the terms of the GNU General Public License as published by
   7  * the Free Software Foundation; either version 2 of the License, or
   8  * (at your option) any later version.
   9  *
  10  * This program is distributed in the hope that it will be useful,
  11  * but WITHOUT ANY WARRANTY; without even the implied warranty of
  12  * MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
  13  * GNU General Public License for more details.
  14  *
  15  * You should have received a copy of the GNU General Public License along
  16  * with this program; if not, write to the Free Software Foundation, Inc.,
  17  * 51 Franklin Street, Fifth Floor, Boston, MA 02110-1301, USA.
  18  * http://www.gnu.org/copyleft/gpl.html
  19  *
  20  * @file
  21  * @ingroup Maintenance
  22  */
  23
  24 use Wikimedia\Rdbms\IDatabase;
  25
  26 require_once __DIR__ . '/Maintenance.php';
  27
  28 /**
  29  * Maintenance script to refresh link tables.
  30  *
  31  * @ingroup Maintenance
  32  */
  33 class RefreshLinks extends Maintenance {
  34         const REPORTING_INTERVAL = 100;
  35
  36         /** @var int|bool */
  37         protected $namespace = false;
  38
  39         public function __construct() {
  40                 parent::__construct();
  41                 $this->addDescription( 'Refresh link tables' );
  42                 $this->addOption( 'dfn-only', 'Delete links from nonexistent articles only' );
  43                 $this->addOption( 'new-only', 'Only affect articles with just a single edit' );
  44                 $this->addOption( 'redirects-only', 'Only fix redirects, not all links' );
  45                 $this->addOption( 'old-redirects-only', 'Only fix redirects with no redirect table entry' );
  46                 $this->addOption( 'e', 'Last page id to refresh', false, true );
  47                 $this->addOption( 'dfn-chunk-size', 'Maximum number of existent IDs to check per ' .
  48                         'query, default 100000', false, true );
  49                 $this->addOption( 'namespace', 'Only fix pages in this namespace', false, true );
  50                 $this->addOption( 'category', 'Only fix pages in this category', false, true );
  51                 $this->addOption( 'tracking-category', 'Only fix pages in this tracking category', false, true );
  52                 $this->addArg( 'start', 'Page_id to start from, default 1', false );
  53                 $this->setBatchSize( 100 );
  54         }
  55
  56         public function execute() {
  57                 // Note that there is a difference between not specifying the start
  58                 // and end IDs and using the minimum and maximum values from the page
  59                 // table. In the latter case, deleteLinksFromNonexistent() will not
  60                 // delete entries for nonexistent IDs that fall outside the range.
  61                 $start = (int)$this->getArg( 0 ) ?: null;
  62                 $end = (int)$this->getOption( 'e' ) ?: null;
  63                 $dfnChunkSize = (int)$this->getOption( 'dfn-chunk-size', 100000 );
  64                 $ns = $this->getOption( 'namespace' );
  65                 if ( $ns === null ) {
  66                         $this->namespace = false;
  67                 } else {
  68                         $this->namespace = (int)$ns;
  69                 }
  70                 if ( ( $category = $this->getOption( 'category', false ) ) !== false ) {
  71                         $title = Title::makeTitleSafe( NS_CATEGORY, $category );
  72                         if ( !$title ) {
  73                                 $this->error( "'$category' is an invalid category name!\n", true );
  74                         }
  75                         $this->refreshCategory( $title );
  76                 } elseif ( ( $category = $this->getOption( 'tracking-category', false ) ) !== false ) {
  77                         $this->refreshTrackingCategory( $category );
  78                 } elseif ( !$this->hasOption( 'dfn-only' ) ) {
  79                         $new = $this->hasOption( 'new-only' );
  80                         $redir = $this->hasOption( 'redirects-only' );
  81                         $oldRedir = $this->hasOption( 'old-redirects-only' );
  82                         $this->doRefreshLinks( $start, $new, $end, $redir, $oldRedir );
  83                         $this->deleteLinksFromNonexistent( null, null, $this->mBatchSize, $dfnChunkSize );
  84                 } else {
  85                         $this->deleteLinksFromNonexistent( $start, $end, $this->mBatchSize, $dfnChunkSize );
  86                 }
  87         }
  88
  89         private function namespaceCond() {
  90                 return $this->namespace !== false
  91                         ? [ 'page_namespace' => $this->namespace ]
  92                         : [];
  93         }
  94
  95         /**
  96          * Do the actual link refreshing.
  97          * @param int|null $start Page_id to start from
  98          * @param bool $newOnly Only do pages with 1 edit
  99          * @param int|null $end Page_id to stop at
 100          * @param bool $redirectsOnly Only fix redirects
 101          * @param bool $oldRedirectsOnly Only fix redirects without redirect entries
 102          */
 103         private function doRefreshLinks( $start, $newOnly = false,
 104                 $end = null, $redirectsOnly = false, $oldRedirectsOnly = false
 105         ) {
 106                 $dbr = $this->getDB( DB_REPLICA, [ 'vslow' ] );
 107
 108                 if ( $start === null ) {
 109                         $start = 1;
 110                 }
 111
 112                 // Give extensions a chance to optimize settings
 113                 Hooks::run( 'MaintenanceRefreshLinksInit', [ $this ] );
 114
 115                 $what = $redirectsOnly ? "redirects" : "links";
 116
 117                 if ( $oldRedirectsOnly ) {
 118                         # This entire code path is cut-and-pasted from below.  Hurrah.
 119
 120                         $conds = [
 121                                 "page_is_redirect=1",
 122                                 "rd_from IS NULL",
 123                                 self::intervalCond( $dbr, 'page_id', $start, $end ),
 124                         ] + $this->namespaceCond();
 125
 126                         $res = $dbr->select(
 127                                 [ 'page', 'redirect' ],
 128                                 'page_id',
 129                                 $conds,
 130                                 __METHOD__,
 131                                 [],
 132                                 [ 'redirect' => [ "LEFT JOIN", "page_id=rd_from" ] ]
 133                         );
 134                         $num = $res->numRows();
 135                         $this->output( "Refreshing $num old redirects from $start...\n" );
 136
 137                         $i = 0;
 138
 139                         foreach ( $res as $row ) {
 140                                 if ( !( ++$i % self::REPORTING_INTERVAL ) ) {
 141                                         $this->output( "$i\n" );
 142                                         wfWaitForSlaves();
 143                                 }
 144                                 $this->fixRedirect( $row->page_id );
 145                         }
 146                 } elseif ( $newOnly ) {
 147                         $this->output( "Refreshing $what from " );
 148                         $res = $dbr->select( 'page',
 149                                 [ 'page_id' ],
 150                                 [
 151                                         'page_is_new' => 1,
 152                                         self::intervalCond( $dbr, 'page_id', $start, $end ),
 153                                 ] + $this->namespaceCond(),
 154                                 __METHOD__
 155                         );
 156                         $num = $res->numRows();
 157                         $this->output( "$num new articles...\n" );
 158
 159                         $i = 0;
 160                         foreach ( $res as $row ) {
 161                                 if ( !( ++$i % self::REPORTING_INTERVAL ) ) {
 162                                         $this->output( "$i\n" );
 163                                         wfWaitForSlaves();
 164                                 }
 165                                 if ( $redirectsOnly ) {
 166                                         $this->fixRedirect( $row->page_id );
 167                                 } else {
 168                                         self::fixLinksFromArticle( $row->page_id, $this->namespace );
 169                                 }
 170                         }
 171                 } else {
 172                         if ( !$end ) {
 173                                 $maxPage = $dbr->selectField( 'page', 'max(page_id)', false );
 174                                 $maxRD = $dbr->selectField( 'redirect', 'max(rd_from)', false );
 175                                 $end = max( $maxPage, $maxRD );
 176                         }
 177                         $this->output( "Refreshing redirects table.\n" );
 178                         $this->output( "Starting from page_id $start of $end.\n" );
 179
 180                         for ( $id = $start; $id <= $end; $id++ ) {
 181                                 if ( !( $id % self::REPORTING_INTERVAL ) ) {
 182                                         $this->output( "$id\n" );
 183                                         wfWaitForSlaves();
 184                                 }
 185                                 $this->fixRedirect( $id );
 186                         }
 187
 188                         if ( !$redirectsOnly ) {
 189                                 $this->output( "Refreshing links tables.\n" );
 190                                 $this->output( "Starting from page_id $start of $end.\n" );
 191
 192                                 for ( $id = $start; $id <= $end; $id++ ) {
 193                                         if ( !( $id % self::REPORTING_INTERVAL ) ) {
 194                                                 $this->output( "$id\n" );
 195                                                 wfWaitForSlaves();
 196                                         }
 197                                         self::fixLinksFromArticle( $id, $this->namespace );
 198                                 }
 199                         }
 200                 }
 201         }
 202
 203         /**
 204          * Update the redirect entry for a given page.
 205          *
 206          * This methods bypasses the "redirect" table to get the redirect target,
 207          * and parses the page's content to fetch it. This allows to be sure that
 208          * the redirect target is up to date and valid.
 209          * This is particularly useful when modifying namespaces to be sure the
 210          * entry in the "redirect" table points to the correct page and not to an
 211          * invalid one.
 212          *
 213          * @param int $id The page ID to check
 214          */
 215         private function fixRedirect( $id ) {
 216                 $page = WikiPage::newFromID( $id );
 217                 $dbw = $this->getDB( DB_MASTER );
 218
 219                 if ( $page === null ) {
 220                         // This page doesn't exist (any more)
 221                         // Delete any redirect table entry for it
 222                         $dbw->delete( 'redirect', [ 'rd_from' => $id ],
 223                                 __METHOD__ );
 224
 225                         return;
 226                 } elseif ( $this->namespace !== false
 227                         && !$page->getTitle()->inNamespace( $this->namespace )
 228                 ) {
 229                         return;
 230                 }
 231
 232                 $rt = null;
 233                 $content = $page->getContent( Revision::RAW );
 234                 if ( $content !== null ) {
 235                         $rt = $content->getUltimateRedirectTarget();
 236                 }
 237
 238                 if ( $rt === null ) {
 239                         // The page is not a redirect
 240                         // Delete any redirect table entry for it
 241                         $dbw->delete( 'redirect', [ 'rd_from' => $id ], __METHOD__ );
 242                         $fieldValue = 0;
 243                 } else {
 244                         $page->insertRedirectEntry( $rt );
 245                         $fieldValue = 1;
 246                 }
 247
 248                 // Update the page table to be sure it is an a consistent state
 249                 $dbw->update( 'page', [ 'page_is_redirect' => $fieldValue ],
 250                         [ 'page_id' => $id ], __METHOD__ );
 251         }
 252
 253         /**
 254          * Run LinksUpdate for all links on a given page_id
 255          * @param int $id The page_id
 256          * @param int|bool $ns Only fix links if it is in this namespace
 257          */
 258         public static function fixLinksFromArticle( $id, $ns = false ) {
 259                 $page = WikiPage::newFromID( $id );
 260
 261                 LinkCache::singleton()->clear();
 262
 263                 if ( $page === null ) {
 264                         return;
 265                 } elseif ( $ns !== false
 266                         && !$page->getTitle()->inNamespace( $ns ) ) {
 267                         return;
 268                 }
 269
 270                 $content = $page->getContent( Revision::RAW );
 271                 if ( $content === null ) {
 272                         return;
 273                 }
 274
 275                 $updates = $content->getSecondaryDataUpdates(
 276                         $page->getTitle(), /* $old = */ null, /* $recursive = */ false );
 277                 foreach ( $updates as $update ) {
 278                         DeferredUpdates::addUpdate( $update );
 279                         DeferredUpdates::doUpdates();
 280                 }
 281         }
 282
 283         /**
 284          * Removes non-existing links from pages from pagelinks, imagelinks,
 285          * categorylinks, templatelinks, externallinks, interwikilinks, langlinks and redirect tables.
 286          *
 287          * @param int|null $start Page_id to start from
 288          * @param int|null $end Page_id to stop at
 289          * @param int $batchSize The size of deletion batches
 290          * @param int $chunkSize Maximum number of existent IDs to check per query
 291          *
 292          * @author Merlijn van Deen <valhallasw@arctus.nl>
 293          */
 294         private function deleteLinksFromNonexistent( $start = null, $end = null, $batchSize = 100,
 295                 $chunkSize = 100000
 296         ) {
 297                 wfWaitForSlaves();
 298                 $this->output( "Deleting illegal entries from the links tables...\n" );
 299                 $dbr = $this->getDB( DB_REPLICA, [ 'vslow' ] );
 300                 do {
 301                         // Find the start of the next chunk. This is based only
 302                         // on existent page_ids.
 303                         $nextStart = $dbr->selectField(
 304                                 'page',
 305                                 'page_id',
 306                                 [ self::intervalCond( $dbr, 'page_id', $start, $end ) ]
 307                                 + $this->namespaceCond(),
 308                                 __METHOD__,
 309                                 [ 'ORDER BY' => 'page_id', 'OFFSET' => $chunkSize ]
 310                         );
 311
 312                         if ( $nextStart !== false ) {
 313                                 // To find the end of the current chunk, subtract one.
 314                                 // This will serve to limit the number of rows scanned in
 315                                 // dfnCheckInterval(), per query, to at most the sum of
 316                                 // the chunk size and deletion batch size.
 317                                 $chunkEnd = $nextStart - 1;
 318                         } else {
 319                                 // This is the last chunk. Check all page_ids up to $end.
 320                                 $chunkEnd = $end;
 321                         }
 322
 323                         $fmtStart = $start !== null ? "[$start" : '(-INF';
 324                         $fmtChunkEnd = $chunkEnd !== null ? "$chunkEnd]" : 'INF)';
 325                         $this->output( "  Checking interval $fmtStart, $fmtChunkEnd\n" );
 326                         $this->dfnCheckInterval( $start, $chunkEnd, $batchSize );
 327
 328                         $start = $nextStart;
 329
 330                 } while ( $nextStart !== false );
 331         }
 332
 333         /**
 334          * @see RefreshLinks::deleteLinksFromNonexistent()
 335          * @param int|null $start Page_id to start from
 336          * @param int|null $end Page_id to stop at
 337          * @param int $batchSize The size of deletion batches
 338          */
 339         private function dfnCheckInterval( $start = null, $end = null, $batchSize = 100 ) {
 340                 $dbw = $this->getDB( DB_MASTER );
 341                 $dbr = $this->getDB( DB_REPLICA, [ 'vslow' ] );
 342
 343                 $linksTables = [ // table name => page_id field
 344                         'pagelinks' => 'pl_from',
 345                         'imagelinks' => 'il_from',
 346                         'categorylinks' => 'cl_from',
 347                         'templatelinks' => 'tl_from',
 348                         'externallinks' => 'el_from',
 349                         'iwlinks' => 'iwl_from',
 350                         'langlinks' => 'll_from',
 351                         'redirect' => 'rd_from',
 352                         'page_props' => 'pp_page',
 353                 ];
 354
 355                 foreach ( $linksTables as $table => $field ) {
 356                         $this->output( "    $table: 0" );
 357                         $tableStart = $start;
 358                         $counter = 0;
 359                         do {
 360                                 $ids = $dbr->selectFieldValues(
 361                                         $table,
 362                                         $field,
 363                                         [
 364                                                 self::intervalCond( $dbr, $field, $tableStart, $end ),
 365                                                 "$field NOT IN ({$dbr->selectSQLText( 'page', 'page_id' )})",
 366                                         ],
 367                                         __METHOD__,
 368                                         [ 'DISTINCT', 'ORDER BY' => $field, 'LIMIT' => $batchSize ]
 369                                 );
 370
 371                                 $numIds = count( $ids );
 372                                 if ( $numIds ) {
 373                                         $counter += $numIds;
 374                                         $dbw->delete( $table, [ $field => $ids ], __METHOD__ );
 375                                         $this->output( ", $counter" );
 376                                         $tableStart = $ids[$numIds - 1] + 1;
 377                                         wfWaitForSlaves();
 378                                 }
 379
 380                         } while ( $numIds >= $batchSize && ( $end === null || $tableStart <= $end ) );
 381
 382                         $this->output( " deleted.\n" );
 383                 }
 384         }
 385
 386         /**
 387          * Build a SQL expression for a closed interval (i.e. BETWEEN).
 388          *
 389          * By specifying a null $start or $end, it is also possible to create
 390          * half-bounded or unbounded intervals using this function.
 391          *
 392          * @param IDatabase $db
 393          * @param string $var Field name
 394          * @param mixed $start First value to include or null
 395          * @param mixed $end Last value to include or null
 396          * @return string
 397          */
 398         private static function intervalCond( IDatabase $db, $var, $start, $end ) {
 399                 if ( $start === null && $end === null ) {
 400                         return "$var IS NOT NULL";
 401                 } elseif ( $end === null ) {
 402                         return "$var >= {$db->addQuotes( $start )}";
 403                 } elseif ( $start === null ) {
 404                         return "$var <= {$db->addQuotes( $end )}";
 405                 } else {
 406                         return "$var BETWEEN {$db->addQuotes( $start )} AND {$db->addQuotes( $end )}";
 407                 }
 408         }
 409
 410         /**
 411          * Refershes links for pages in a tracking category
 412          *
 413          * @param string $category Category key
 414          */
 415         private function refreshTrackingCategory( $category ) {
 416                 $cats = $this->getPossibleCategories( $category );
 417
 418                 if ( !$cats ) {
 419                         $this->error( "Tracking category '$category' is disabled\n" );
 420                         // Output to stderr but don't bail out,
 421                 }
 422
 423                 foreach ( $cats as $cat ) {
 424                         $this->refreshCategory( $cat );
 425                 }
 426         }
 427
 428         /**
 429          * Refreshes links to a category
 430          *
 431          * @param Title $category
 432          */
 433         private function refreshCategory( Title $category ) {
 434                 $this->output( "Refreshing pages in category '{$category->getText()}'...\n" );
 435
 436                 $dbr = $this->getDB( DB_REPLICA );
 437                 $conds = [
 438                         'page_id=cl_from',
 439                         'cl_to' => $category->getDBkey(),
 440                 ];
 441                 if ( $this->namespace !== false ) {
 442                         $conds['page_namespace'] = $this->namespace;
 443                 }
 444
 445                 $i = 0;
 446                 $timestamp = '';
 447                 $lastId = 0;
 448                 do {
 449                         $finalConds = $conds;
 450                         $timestamp = $dbr->addQuotes( $timestamp );
 451                         $finalConds [] =
 452                                 "(cl_timestamp > $timestamp OR (cl_timestamp = $timestamp AND cl_from > $lastId))";
 453                         $res = $dbr->select( [ 'page', 'categorylinks' ],
 454                                 [ 'page_id', 'cl_timestamp' ],
 455                                 $finalConds,
 456                                 __METHOD__,
 457                                 [
 458                                         'ORDER BY' => [ 'cl_timestamp', 'cl_from' ],
 459                                         'LIMIT' => $this->mBatchSize,
 460                                 ]
 461                         );
 462
 463                         foreach ( $res as $row ) {
 464                                 if ( !( ++$i % self::REPORTING_INTERVAL ) ) {
 465                                         $this->output( "$i\n" );
 466                                         wfWaitForSlaves();
 467                                 }
 468                                 $lastId = $row->page_id;
 469                                 $timestamp = $row->cl_timestamp;
 470                                 self::fixLinksFromArticle( $row->page_id );
 471                         }
 472
 473                 } while ( $res->numRows() == $this->mBatchSize );
 474         }
 475
 476         /**
 477          * Returns a list of possible categories for a given tracking category key
 478          *
 479          * @param string $categoryKey
 480          * @return Title[]
 481          */
 482         private function getPossibleCategories( $categoryKey ) {
 483                 $trackingCategories = new TrackingCategories( $this->getConfig() );
 484                 $cats = $trackingCategories->getTrackingCategories();
 485                 if ( isset( $cats[$categoryKey] ) ) {
 486                         return $cats[$categoryKey]['cats'];
 487                 }
 488                 $this->error( "Unknown tracking category {$categoryKey}\n", true );
 489         }
 490 }
 491
 492 $maintClass = 'RefreshLinks';
 493 require_once RUN_MAINTENANCE_IF_MAIN;