dumpCategoriesAsRdf.php

Go to the documentation of this file.

<?php
use MediaWiki\Category\CategoriesRdf;
use MediaWiki\MainConfigNames;
use MediaWiki\Maintenance\Maintenance;
use Wikimedia\Purtle\RdfWriter;
use Wikimedia\Purtle\RdfWriterFactory;
use Wikimedia\Rdbms\IReadableDatabase;
 
// @codeCoverageIgnoreStart
require_once __DIR__ . '/Maintenance.php';
// @codeCoverageIgnoreEnd
 
class DumpCategoriesAsRdf extends Maintenance {
    private $rdfWriter;
    private $categoriesRdf;
 
    public function __construct() {
        parent::__construct();
 
        $this->addDescription( "Generate RDF dump of categories in a wiki." );
 
        $this->setBatchSize( 200 );
        $this->addOption( 'output', "Output file (default is stdout). Will be overwritten.",
            false, true );
        $this->addOption( 'format', "Set the dump format.", false, true );
    }
 
    public function getCategoryIterator( IReadableDatabase $dbr, $fname ) {
        $it = new BatchRowIterator(
            $dbr,
            $dbr->newSelectQueryBuilder()
                ->from( 'page' )
                ->leftJoin( 'page_props', null, [ 'pp_propname' => 'hiddencat', 'pp_page = page_id' ] )
                ->leftJoin( 'category', null, [ 'cat_title = page_title' ] )
                ->select( [
                    'page_title',
                    'page_id',
                    'pp_propname',
                    'cat_pages',
                    'cat_subcats',
                    'cat_files'
                ] )
                ->where( [
                    'page_namespace' => NS_CATEGORY,
                ] )
                ->caller( $fname ),
            [ 'page_title' ],
            $this->getBatchSize()
        );
        return $it;
    }
 
    public 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->getBatchSize()
        );
        return new RecursiveIteratorIterator( $it );
    }
 
    public function addDumpHeader( $timestamp ) {
        $licenseUrl = $this->getConfig()->get( MainConfigNames::RightsUrl );
        if ( str_starts_with( $licenseUrl, '//' ) ) {
            $licenseUrl = 'https:' . $licenseUrl;
        }
        $urlUtils = $this->getServiceContainer()->getUrlUtils();
        $this->rdfWriter->about( $this->categoriesRdf->getDumpURI() )
            ->a( 'schema', 'Dataset' )
            ->a( 'owl', 'Ontology' )
            ->say( 'cc', 'license' )->is( $licenseUrl )
            ->say( 'schema', 'softwareVersion' )->value( CategoriesRdf::FORMAT_VERSION )
            ->say( 'schema', 'dateModified' )
            ->value( wfTimestamp( TS_ISO_8601, $timestamp ), 'xsd', 'dateTime' )
            ->say( 'schema', 'isPartOf' )->is( (string)$urlUtils->expand( '/', PROTO_CANONICAL ) )
            ->say( 'owl', 'imports' )->is( CategoriesRdf::OWL_URL );
    }
 
    public function execute() {
        $outFile = $this->getOption( 'output', 'php://stdout' );
 
        if ( $outFile === '-' ) {
            $outFile = 'php://stdout';
        }
 
        $output = fopen( $outFile, 'w' );
        $this->rdfWriter = $this->createRdfWriter( $this->getOption( 'format', 'ttl' ) );
        $this->categoriesRdf = new CategoriesRdf( $this->rdfWriter );
 
        $this->categoriesRdf->setupPrefixes();
        $this->rdfWriter->start();
 
        $this->addDumpHeader( time() );
        fwrite( $output, $this->rdfWriter->drain() );
 
        $dbr = $this->getDB( DB_REPLICA, [ 'vslow' ] );
 
        foreach ( $this->getCategoryIterator( $dbr, __METHOD__ ) as $batch ) {
            $pages = [];
            foreach ( $batch as $row ) {
                $this->categoriesRdf->writeCategoryData(
                    $row->page_title,
                    $row->pp_propname === 'hiddencat',
                    (int)$row->cat_pages - (int)$row->cat_subcats - (int)$row->cat_files,
                    (int)$row->cat_subcats
                );
                if ( $row->page_id ) {
                    $pages[$row->page_id] = $row->page_title;
                }
            }
 
            foreach ( $this->getCategoryLinksIterator( $dbr, array_keys( $pages ), __METHOD__ ) as $row ) {
                $this->categoriesRdf->writeCategoryLinkData( $pages[$row->cl_from], $row->cl_to );
            }
            fwrite( $output, $this->rdfWriter->drain() );
        }
        fflush( $output );
        if ( $outFile !== '-' ) {
            fclose( $output );
        }
    }
 
    private function createRdfWriter( $format ) {
        $factory = new RdfWriterFactory();
        return $factory->getWriter( $factory->getFormatName( $format ) );
    }
}
 
// @codeCoverageIgnoreStart
$maintClass = DumpCategoriesAsRdf::class;
require_once RUN_MAINTENANCE_IF_MAIN;
// @codeCoverageIgnoreEnd