SearchUpdate.php

Go to the documentation of this file.

<?php
class SearchUpdate implements DeferrableUpdate {
    private $id = 0;

    private $title;

    private $content;

    public function __construct( $id, $title, $c = false ) {
        if ( is_string( $title ) ) {
            $nt = Title::newFromText( $title );
        } else {
            $nt = $title;
        }

        if ( $nt ) {
            $this->id = $id;
            // is_string() check is back-compat for ApprovedRevs
            if ( is_string( $c ) ) {
                $this->content = new TextContent( $c );
            } else {
                $this->content = $c ?: false;
            }
            $this->title = $nt;
        } else {
            wfDebug( "SearchUpdate object created with invalid title '$title'\n" );
        }
    }

    public function doUpdate() {
        global $wgDisableSearchUpdate;

        if ( $wgDisableSearchUpdate || !$this->id ) {
            return;
        }

        wfProfileIn( __METHOD__ );

        $page = WikiPage::newFromId( $this->id, WikiPage::READ_LATEST );

        foreach ( SearchEngine::getSearchTypes() as $type ) {
            $search = SearchEngine::create( $type );
            $indexTitle = $this->indexTitle( $search );
            if ( !$search->supports( 'search-update' ) ) {
                continue;
            }

            $normalTitle = $search->normalizeText( $indexTitle );

            if ( $page === null ) {
                $search->delete( $this->id, $normalTitle );
                continue;
            } elseif ( $this->content === false ) {
                $search->updateTitle( $this->id, $normalTitle );
                continue;
            }

            $text = $search->getTextFromContent( $this->title, $this->content );
            if ( !$search->textAlreadyUpdatedForIndex() ) {
                $text = self::updateText( $text );
            }

            # Perform the actual update
            $search->update( $this->id, $normalTitle, $search->normalizeText( $text ) );
        }

        wfProfileOut( __METHOD__ );
    }

    public static function updateText( $text ) {
        global $wgContLang;

        # Language-specific strip/conversion
        $text = $wgContLang->normalizeForSearch( $text );
        $lc = SearchEngine::legalSearchChars() . '&#;';

        wfProfileIn( __METHOD__ . '-regexps' );
        $text = preg_replace( "/<\\/?\\s*[A-Za-z][^>]*?>/",
            ' ', $wgContLang->lc( " " . $text . " " ) ); # Strip HTML markup
        $text = preg_replace( "/(^|\\n)==\\s*([^\\n]+)\\s*==(\\s)/sD",
            "\\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 );

        # Internal image links
        $pat2 = "/\\[\\[image:([{$uc}]+)\\.(gif|png|jpg|jpeg)([^{$uc}])/i";
        $text = preg_replace( $pat2, " \\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 );

        # Handle 's, s'
        #
        #   $text = preg_replace( "/([{$lc}]+)'s /", "\\1 \\1's ", $text );
        #   $text = preg_replace( "/([{$lc}]+)s' /", "\\1s ", $text );
        #
        # These tail-anchored regexps are insanely slow. The worst case comes
        # when Japanese or Chinese text (ie, no word spacing) is written on
        # a wiki configured for Western UTF-8 mode. The Unicode characters are
        # expanded to hex codes and the "words" are very long paragraph-length
        # monstrosities. On a large page the above regexps may take over 20
        # seconds *each* on a 1GHz-level processor.
        #
        # Following are reversed versions which are consistently fast
        # (about 3 milliseconds on 1GHz-level processor).
        #
        $text = strrev( preg_replace( "/ s'([{$lc}]+)/", " s'\\1 \\1", strrev( $text ) ) );
        $text = strrev( preg_replace( "/ 's([{$lc}]+)/", " s\\1", strrev( $text ) ) );

        # Strip wiki '' and '''
        $text = preg_replace( "/''[']*/", " ", $text );
        wfProfileOut( __METHOD__ . '-regexps' );

        return $text;
    }

    private function indexTitle( SearchEngine $search ) {
        global $wgContLang;

        $ns = $this->title->getNamespace();
        $title = $this->title->getText();

        $lc = $search->legalSearchChars() . '&#;';
        $t = $wgContLang->normalizeForSearch( $title );
        $t = preg_replace( "/[^{$lc}]+/", ' ', $t );
        $t = $wgContLang->lc( $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 );

        if ( $ns == NS_FILE ) {
            $t = preg_replace( "/ (png|gif|jpg|jpeg|ogg)$/", "", $t );
        }
        return trim( $t );
    }
}