categoryChangesAsRdf.php

Go to the documentation of this file.

<?php
use MediaWiki\Category\CategoriesRdf;
use MediaWiki\MainConfigNames;
use MediaWiki\Maintenance\Maintenance;
use MediaWiki\Utils\MWTimestamp;
use Wikimedia\Purtle\RdfWriter;
use Wikimedia\Purtle\TurtleRdfWriter;
use Wikimedia\Rdbms\IReadableDatabase;
 
// @codeCoverageIgnoreStart
require_once __DIR__ . '/Maintenance.php';
// @codeCoverageIgnoreEnd
 
class CategoryChangesAsRdf extends Maintenance {
    private const SPARQL_INSERT = <<<SPARQL
INSERT DATA {
%s
};
 
SPARQL;
 
    private const SPARQL_DELETE = <<<SPARQLD
DELETE {
?category ?x ?y
} WHERE {
   ?category ?x ?y
   VALUES ?category {
     %s
   }
};
 
SPARQLD;
 
    private $rdfWriter;
    private $categoriesRdf;
 
    private $startTS;
    private $endTS;
 
    protected $processed = [];
 
    public function __construct() {
        parent::__construct();
 
        $this->addDescription( "Generate RDF dump of category changes in a wiki." );
 
        $this->setBatchSize( 200 );
        $this->addOption( 'output', "Output file (default is stdout). Will be overwritten.", false,
            true, 'o' );
        $this->addOption( 'start', 'Starting timestamp (inclusive), in ISO or MediaWiki format.',
            true, true, 's' );
        $this->addOption( 'end', 'Ending timestamp (exclusive), in ISO or MediaWiki format.', true,
            true, 'e' );
    }
 
    public function initialize() {
        // SPARQL Update syntax is close to Turtle format, so we can use Turtle writer.
        $this->rdfWriter = new TurtleRdfWriter();
        $this->categoriesRdf = new CategoriesRdf( $this->rdfWriter );
    }
 
    public function execute() {
        $this->initialize();
        $startTS = new MWTimestamp( $this->getOption( "start" ) );
 
        $endTS = new MWTimestamp( $this->getOption( "end" ) );
        $now = new MWTimestamp();
        $rcMaxAge = $this->getConfig()->get( MainConfigNames::RCMaxAge );
 
        if ( (int)$now->getTimestamp( TS_UNIX ) - (int)$startTS->getTimestamp( TS_UNIX ) > $rcMaxAge ) {
            $this->error( "Start timestamp too old, maximum RC age is $rcMaxAge!" );
        }
        if ( (int)$now->getTimestamp( TS_UNIX ) - (int)$endTS->getTimestamp( TS_UNIX ) > $rcMaxAge ) {
            $this->error( "End timestamp too old, maximum RC age is $rcMaxAge!" );
        }
 
        $this->startTS = $startTS->getTimestamp();
        $this->endTS = $endTS->getTimestamp();
 
        $outFile = $this->getOption( 'output', 'php://stdout' );
        if ( $outFile === '-' ) {
            $outFile = 'php://stdout';
        }
 
        $output = fopen( $outFile, 'wb' );
 
        $this->categoriesRdf->setupPrefixes();
        $this->rdfWriter->start();
 
        $prefixes = $this->getRdf();
        // We have to strip @ from prefix, since SPARQL UPDATE doesn't use them
        // Also strip dot at the end.
        $prefixes = preg_replace( [ '/^@/m', '/\s*[.]$/m' ], '', $prefixes );
        fwrite( $output, $prefixes );
 
        $dbr = $this->getDB( DB_REPLICA, [ 'vslow' ] );
 
        // Deletes go first because if the page was deleted, other changes
        // do not matter. This only gets true deletes, i.e. not pages that were restored.
        $this->handleDeletes( $dbr, $output );
        // Moves go before additions because if category is moved, we should not process creation
        // as it would produce wrong data - because create row has old title
        $this->handleMoves( $dbr, $output );
        // We need to handle restores too since delete may have happened in previous update.
        $this->handleRestores( $dbr, $output );
        // Process newly added pages
        $this->handleAdds( $dbr, $output );
        // Process page edits
        $this->handleEdits( $dbr, $output );
        // Process categorization changes
        $this->handleCategorization( $dbr, $output );
 
        // Update timestamp
        fwrite( $output, $this->updateTS( $this->endTS ) );
    }
 
    private function getInsertRdf() {
        $rdfText = $this->getRdf();
        if ( !$rdfText ) {
            return "";
        }
        return sprintf( self::SPARQL_INSERT, $rdfText );
    }
 
    private function getCategoriesUpdate( IReadableDatabase $dbr, $deleteUrls, $pages, $mark ) {
        if ( !$deleteUrls ) {
            return "";
        }
 
        if ( $pages ) {
            $this->writeParentCategories( $dbr, $pages );
        }
 
        return "# $mark\n" . sprintf( self::SPARQL_DELETE, implode( ' ', $deleteUrls ) ) .
            $this->getInsertRdf();
    }
 
    private function writeParentCategories( IReadableDatabase $dbr, $pages ) {
        foreach ( $this->getCategoryLinksIterator( $dbr, array_keys( $pages ), __METHOD__ ) as $row ) {
            $this->categoriesRdf->writeCategoryLinkData( $pages[$row->cl_from], $row->cl_to );
        }
    }
 
    public function updateTS( $timestamp ) {
        $dumpUrl = '<' . $this->categoriesRdf->getDumpURI() . '>';
        $ts = wfTimestamp( TS_ISO_8601, $timestamp );
        $tsQuery = <<<SPARQL
DELETE {
  $dumpUrl schema:dateModified ?o .
}
WHERE {
  $dumpUrl schema:dateModified ?o .
};
INSERT DATA {
  $dumpUrl schema:dateModified "$ts"^^xsd:dateTime .
}
 
SPARQL;
        return $tsQuery;
    }
 
    private function setupChangesIterator(
        IReadableDatabase $dbr,
        array $columns,
        string $fname
    ) {
        $it = new BatchRowIterator( $dbr,
            $dbr->newSelectQueryBuilder()
                ->from( 'recentchanges' )
                ->leftJoin( 'page_props', null, [ 'pp_propname' => 'hiddencat', 'pp_page = rc_cur_id' ] )
                ->leftJoin( 'category', null, [ 'cat_title = rc_title' ] )
                ->select( array_merge( $columns, [
                    'rc_title',
                    'rc_cur_id',
                    'pp_propname',
                    'cat_pages',
                    'cat_subcats',
                    'cat_files'
                ] ) )
                ->caller( $fname ),
            [ 'rc_timestamp' ],
            $this->mBatchSize
        );
        $this->addTimestampConditions( $it, $dbr );
        return $it;
    }
 
    protected function getNewCatsIterator( IReadableDatabase $dbr, $fname ) {
        $it = $this->setupChangesIterator( $dbr, [], $fname );
        $it->sqb->conds( [
            'rc_namespace' => NS_CATEGORY,
            'rc_new' => 1,
        ] );
        return $it;
    }
 
    protected function getMovedCatsIterator( IReadableDatabase $dbr, $fname ) {
        $it = $this->setupChangesIterator(
            $dbr,
            [ 'page_title', 'page_namespace' ],
            $fname
        );
        $it->sqb->conds( [
            'rc_namespace' => NS_CATEGORY,
            'rc_new' => 0,
            'rc_log_type' => 'move',
            'rc_type' => RC_LOG,
        ] );
        $it->sqb->join( 'page', null, 'rc_cur_id = page_id' );
        $this->addIndex( $it );
        return $it;
    }
 
    protected function getDeletedCatsIterator( IReadableDatabase $dbr, $fname ) {
        $it = new BatchRowIterator( $dbr,
            $dbr->newSelectQueryBuilder()
                ->from( 'recentchanges' )
                ->select( [ 'rc_cur_id', 'rc_title' ] )
                ->where( [
                    'rc_namespace' => NS_CATEGORY,
                    'rc_new' => 0,
                    'rc_log_type' => 'delete',
                    'rc_log_action' => 'delete',
                    'rc_type' => RC_LOG,
                    // We will fetch ones that do not have page record. If they do,
                    // this means they were restored, thus restoring handler will pick it up.
                    'NOT EXISTS (SELECT * FROM page WHERE page_id = rc_cur_id)',
                ] )
                ->caller( $fname ),
            [ 'rc_timestamp' ],
            $this->mBatchSize
        );
        $this->addTimestampConditions( $it, $dbr );
        $this->addIndex( $it );
        return $it;
    }
 
    protected function getRestoredCatsIterator( IReadableDatabase $dbr, $fname ) {
        $it = $this->setupChangesIterator( $dbr, [], $fname );
        $it->sqb->conds( [
            'rc_namespace' => NS_CATEGORY,
            'rc_new' => 0,
            'rc_log_type' => 'delete',
            'rc_log_action' => 'restore',
            'rc_type' => RC_LOG,
            // We will only fetch ones that have page record
            'EXISTS (SELECT page_id FROM page WHERE page_id = rc_cur_id)',
        ] );
        $this->addIndex( $it );
        return $it;
    }
 
    protected function getChangedCatsIterator( IReadableDatabase $dbr, $type, $fname ) {
        $it = $this->setupChangesIterator( $dbr, [], $fname );
        $it->sqb->conds( [
            'rc_namespace' => NS_CATEGORY,
            'rc_new' => 0,
            'rc_type' => $type,
        ] );
        $this->addIndex( $it );
        return $it;
    }
 
    private function addTimestampConditions( BatchRowIterator $it, IReadableDatabase $dbr ) {
        $it->sqb->conds( [
            $dbr->expr( 'rc_timestamp', '>=', $dbr->timestamp( $this->startTS ) ),
            $dbr->expr( 'rc_timestamp', '<', $dbr->timestamp( $this->endTS ) ),
        ] );
    }
 
    private function addIndex( BatchRowIterator $it ) {
        $it->sqb->options( [
            'USE INDEX' => [ 'recentchanges' => 'rc_new_name_timestamp' ]
        ] );
    }
 
    protected function getCategoryLinksIterator( IReadableDatabase $dbr, array $ids, $fname ) {
        $it = new BatchRowIterator(
            $dbr,
            $dbr->newSelectQueryBuilder()
                ->from( 'categorylinks' )
                ->select( [ 'cl_from', 'cl_to' ] )
                ->where( [
                    'cl_type' => 'subcat',
                    'cl_from' => $ids
                ] )
                ->caller( $fname ),
            [ 'cl_from', 'cl_to' ],
            $this->mBatchSize
        );
        return new RecursiveIteratorIterator( $it );
    }
 
    public function getRdf() {
        return $this->rdfWriter->drain();
    }
 
    public function handleDeletes( IReadableDatabase $dbr, $output ) {
        // This only does "true" deletes - i.e. those that the page stays deleted
 
        foreach ( $this->getDeletedCatsIterator( $dbr, __METHOD__ ) as $batch ) {
            $deleteUrls = [];
            foreach ( $batch as $row ) {
                // This can produce duplicates, we don't care
                $deleteUrls[] = '<' . $this->categoriesRdf->labelToUrl( $row->rc_title ) . '>';
                $this->processed[$row->rc_cur_id] = true;
            }
            fwrite( $output, $this->getCategoriesUpdate( $dbr, $deleteUrls, [], "Deletes" ) );
        }
    }
 
    private function writeCategoryData( $row ) {
        $this->categoriesRdf->writeCategoryData(
            $row->rc_title,
            $row->pp_propname === 'hiddencat',
            (int)$row->cat_pages - (int)$row->cat_subcats - (int)$row->cat_files,
            (int)$row->cat_subcats
        );
    }
 
    public function handleMoves( IReadableDatabase $dbr, $output ) {
        foreach ( $this->getMovedCatsIterator( $dbr, __METHOD__ ) as $batch ) {
            $pages = [];
            $deleteUrls = [];
            foreach ( $batch as $row ) {
                $deleteUrls[] = '<' . $this->categoriesRdf->labelToUrl( $row->rc_title ) . '>';
 
                if ( isset( $this->processed[$row->rc_cur_id] ) ) {
                    // We already captured this one before
                    continue;
                }
 
                if ( $row->page_namespace != NS_CATEGORY ) {
                    // If page was moved out of Category:, we'll just delete
                    continue;
                }
                $row->rc_title = $row->page_title;
                $this->writeCategoryData( $row );
                $pages[$row->rc_cur_id] = $row->page_title;
                $this->processed[$row->rc_cur_id] = true;
            }
 
            fwrite( $output, $this->getCategoriesUpdate( $dbr, $deleteUrls, $pages, "Moves" ) );
        }
    }
 
    public function handleRestores( IReadableDatabase $dbr, $output ) {
        fwrite( $output, "# Restores\n" );
 
        // This will only find those restores that were not deleted later.
        foreach ( $this->getRestoredCatsIterator( $dbr, __METHOD__ ) as $batch ) {
            $pages = [];
            foreach ( $batch as $row ) {
                if ( isset( $this->processed[$row->rc_cur_id] ) ) {
                    // We already captured this one before
                    continue;
                }
                $this->writeCategoryData( $row );
                $pages[$row->rc_cur_id] = $row->rc_title;
                $this->processed[$row->rc_cur_id] = true;
            }
 
            if ( !$pages ) {
                continue;
            }
 
            $this->writeParentCategories( $dbr, $pages );
 
            fwrite( $output, $this->getInsertRdf() );
        }
    }
 
    public function handleAdds( IReadableDatabase $dbr, $output ) {
        fwrite( $output, "# Additions\n" );
 
        foreach ( $this->getNewCatsIterator( $dbr, __METHOD__ ) as $batch ) {
            $pages = [];
            foreach ( $batch as $row ) {
                if ( isset( $this->processed[$row->rc_cur_id] ) ) {
                    // We already captured this one before
                    continue;
                }
                $this->writeCategoryData( $row );
                $pages[$row->rc_cur_id] = $row->rc_title;
                $this->processed[$row->rc_cur_id] = true;
            }
 
            if ( !$pages ) {
                continue;
            }
 
            $this->writeParentCategories( $dbr, $pages );
            fwrite( $output, $this->getInsertRdf() );
        }
    }
 
    public function handleEdits( IReadableDatabase $dbr, $output ) {
        // Editing category can change hidden flag and add new parents.
        // TODO: it's pretty expensive to update all edited categories, and most edits
        // aren't actually interesting for us. Some way to know which are interesting?
        // We can capture recategorization on the next step, but not change in hidden status.
 
        foreach ( $this->getChangedCatsIterator( $dbr, RC_EDIT, __METHOD__ ) as $batch ) {
            $pages = [];
            $deleteUrls = [];
            foreach ( $batch as $row ) {
                // Note that on categorization event, cur_id points to
                // the child page, not the parent category!
                if ( isset( $this->processed[$row->rc_cur_id] ) ) {
                    // We already captured this one before
                    continue;
                }
                $this->writeCategoryData( $row );
                $pages[$row->rc_cur_id] = $row->rc_title;
                $this->processed[$row->rc_cur_id] = true;
                $deleteUrls[] = '<' . $this->categoriesRdf->labelToUrl( $row->rc_title ) . '>';
            }
 
            fwrite( $output, $this->getCategoriesUpdate( $dbr, $deleteUrls, $pages, "Edits" ) );
        }
    }
 
    public function handleCategorization( IReadableDatabase $dbr, $output ) {
        $processedTitle = [];
 
        // Categorization change can add new parents and change counts
        // for the parent category.
 
        foreach ( $this->getChangedCatsIterator( $dbr, RC_CATEGORIZE, __METHOD__ ) as $batch ) {
            /*
             * Note that on categorization event, cur_id points to
             * the child page, not the parent category!
             * So we need to have a two-stage process, since we have ID from one
             * category and title from another, and we need both for proper updates.
             * TODO: For now, we do full update even though some data hasn't changed,
             * e.g. parents for parent cat and counts for child cat.
             */
            $childPages = [];
            $parentCats = [];
            foreach ( $batch as $row ) {
                $childPages[$row->rc_cur_id] = true;
                $parentCats[$row->rc_title] = true;
            }
 
            $pages = [];
            $deleteUrls = [];
 
            if ( $childPages ) {
                // Load child rows by ID
                $childRows = $dbr->newSelectQueryBuilder()
                    ->select( [
                        'page_id',
                        'rc_title' => 'page_title',
                        'pp_propname',
                        'cat_pages',
                        'cat_subcats',
                        'cat_files',
                    ] )
                    ->from( 'page' )
                    ->leftJoin( 'page_props', null, [ 'pp_propname' => 'hiddencat', 'pp_page = page_id' ] )
                    ->leftJoin( 'category', null, [ 'cat_title = page_title' ] )
                    ->where( [ 'page_namespace' => NS_CATEGORY, 'page_id' => array_keys( $childPages ) ] )
                    ->caller( __METHOD__ )->fetchResultSet();
                foreach ( $childRows as $row ) {
                    if ( isset( $this->processed[$row->page_id] ) ) {
                        // We already captured this one before
                        continue;
                    }
                    $this->writeCategoryData( $row );
                    if ( $row->page_id ) {
                        $pages[$row->page_id] = $row->rc_title;
                        $deleteUrls[] = '<' . $this->categoriesRdf->labelToUrl( $row->rc_title ) . '>';
                        $this->processed[$row->page_id] = true;
                    }
                }
            }
 
            if ( $parentCats ) {
                // Load parent rows by title
                $parentRows = $dbr->newSelectQueryBuilder()
                    ->select( [
                        'page_id',
                        'rc_title' => 'cat_title',
                        'pp_propname',
                        'cat_pages',
                        'cat_subcats',
                        'cat_files',
                    ] )
                    ->from( 'category' )
                    ->leftJoin( 'page', null, [ 'page_title = cat_title', 'page_namespace' => NS_CATEGORY ] )
                    ->leftJoin( 'page_props', null, [ 'pp_propname' => 'hiddencat', 'pp_page = page_id' ] )
                    ->where( [ 'cat_title' => array_map( 'strval', array_keys( $parentCats ) ) ] )
                    ->caller( __METHOD__ )->fetchResultSet();
                foreach ( $parentRows as $row ) {
                    if ( $row->page_id && isset( $this->processed[$row->page_id] ) ) {
                        // We already captured this one before
                        continue;
                    }
                    if ( isset( $processedTitle[$row->rc_title] ) ) {
                        // We already captured this one before
                        continue;
                    }
                    $this->writeCategoryData( $row );
                    if ( $row->page_id ) {
                        $pages[$row->page_id] = $row->rc_title;
                        $deleteUrls[] = '<' . $this->categoriesRdf->labelToUrl( $row->rc_title ) . '>';
                        $this->processed[$row->page_id] = true;
                    }
                    $processedTitle[$row->rc_title] = true;
                }
            }
 
            fwrite( $output, $this->getCategoriesUpdate( $dbr, $deleteUrls, $pages, "Changes" ) );
        }
    }
}
 
// @codeCoverageIgnoreStart
$maintClass = CategoryChangesAsRdf::class;
require_once RUN_MAINTENANCE_IF_MAIN;
// @codeCoverageIgnoreEnd