SearchUpdate.php

Go to the documentation of this file.

<?php
namespace MediaWiki\Search;
 
use MediaWiki\Content\Content;
use MediaWiki\Logger\LoggerFactory;
use MediaWiki\MainConfigNames;
use MediaWiki\MediaWikiServices;
use MediaWiki\Page\ExistingPageRecord;
use MediaWiki\Page\PageIdentity;
use MediaWiki\Title\Title;
use SearchEngine;
use Wikimedia\Rdbms\IDBAccessObject;
 
class SearchUpdate {
    private $id = 0;
 
    private $page;
 
    private $content;
 
    private $latestPage = null;
 
    public function __construct( $id, $page, ?Content $c = null ) {
        $this->page = $page;
        $this->id = $id;
        $this->content = $c;
    }
 
    public function doUpdate() {
        $services = MediaWikiServices::getInstance();
        $searchEngineConfig = $services->getSearchEngineConfig();
 
        if ( $services->getMainConfig()->get( MainConfigNames::DisableSearchUpdate ) || !$this->id ) {
            LoggerFactory::getInstance( "search" )
                ->debug( "Skipping update: search updates disabled by config" );
            return;
        }
 
        $seFactory = $services->getSearchEngineFactory();
        foreach ( $searchEngineConfig->getSearchTypes() as $type ) {
            $search = $seFactory->create( $type );
            if ( !$search->supports( 'search-update' ) ) {
                continue;
            }
 
            $normalTitle = $this->getNormalizedTitle( $search );
 
            if ( $this->getLatestPage() === null ) {
                $search->delete( $this->id, $normalTitle );
                continue;
            }
            if ( $this->content === null ) {
                $search->updateTitle( $this->id, $normalTitle );
                continue;
            }
 
            $text = $this->content->getTextForSearchIndex();
            $text = $this->updateText( $text, $search );
 
            # Perform the actual update
            $search->update( $this->id, $normalTitle, $search->normalizeText( $text ) );
        }
    }
 
    public function updateText( $text, ?SearchEngine $se = null ) {
        $services = MediaWikiServices::getInstance();
        $contLang = $services->getContentLanguage();
        # Language-specific strip/conversion
        $text = $contLang->normalizeForSearch( $text );
        $se = $se ?: $services->newSearchEngine();
        $lc = $se->legalSearchChars() . '&#;';
 
        # Strip HTML markup
        $text = preg_replace( "/<\\/?\\s*[A-Za-z][^>]*?>/",
            ' ', $contLang->lc( " " . $text . " " ) );
        $text = preg_replace( "/(^|\\n)==\\s*([^\\n]+)\\s*==(\\s)/",
            "\\1\\2 \\2 \\2\\3", $text ); # Emphasize headings
 
        # Strip external URLs
        $uc = "A-Za-z0-9_\\/:.,~%\\-+&;#?!=()@\\x80-\\xFF";
        $protos = "http|https|ftp|mailto|news|gopher";
        $pat = "/(^|[^\\[])({$protos}):[{$uc}]+([^{$uc}]|$)/";
        $text = preg_replace( $pat, "\\1 \\3", $text );
 
        $p1 = "/([^\\[])\\[({$protos}):[{$uc}]+]/";
        $p2 = "/([^\\[])\\[({$protos}):[{$uc}]+\\s+([^\\]]+)]/";
        $text = preg_replace( $p1, "\\1 ", $text );
        $text = preg_replace( $p2, "\\1 \\3 ", $text );
 
        $text = preg_replace( "/([^{$lc}])([{$lc}]+)]]([a-z]+)/",
            "\\1\\2 \\2\\3", $text ); # Handle [[game]]s
 
        # Strip all remaining non-search characters
        $text = preg_replace( "/[^{$lc}]+/", " ", $text );
 
        $text = strrev( preg_replace( "/ s'([{$lc}]+)/", " s'\\1 \\1", strrev( $text ) ) );
        $text = strrev( preg_replace( "/ 's([{$lc}]+)/", " s\\1", strrev( $text ) ) );
 
        # Strip wiki '' and '''
        $text = preg_replace( "/''[']*/", " ", $text );
 
        return $text;
    }
 
    private function getLatestPage() {
        if ( !$this->latestPage ) {
            $this->latestPage = MediaWikiServices::getInstance()->getPageStore()
                ->getPageById( $this->id, IDBAccessObject::READ_LATEST );
        }
 
        return $this->latestPage;
    }
 
    private function getNormalizedTitle( SearchEngine $search ) {
        $contLang = MediaWikiServices::getInstance()->getContentLanguage();
        $title = Title::newFromPageIdentity( $this->page )->getText();
 
        $lc = $search->legalSearchChars() . '&#;';
        $t = $contLang->normalizeForSearch( $title );
        $t = preg_replace( "/[^{$lc}]+/", ' ', $t );
        $t = $contLang->lc( $t );
 
        if ( $this->page->getNamespace() === NS_FILE ) {
            $t = preg_replace( "/([{$lc}]+)\\.(\\w{1,4})$/", "\\1 \\1.\\2", $t );
        }
 
        # Handle 's, s'
        $t = preg_replace( "/([{$lc}]+)'s( |$)/", "\\1 \\1's ", $t );
        $t = preg_replace( "/([{$lc}]+)s'( |$)/", "\\1s ", $t );
 
        $t = preg_replace( "/\\s+/", ' ', $t );
 
        return $search->normalizeText( trim( $t ) );
    }
}