Sven Hartrumpf wrote:

Hi.

A question about dumpHTML.inc:

    function getFriendlyName( $name ) {
            global $wgLang;
            # Replace illegal characters for Windows paths with underscores
            $friendlyName = strtr( $name, '/\\*?"<>|~', '_________' );

            # Work out lower case form. We assume we're on a system with case-insensitive
            # filenames, so unless the case is of a special form, we have to disambiguate
            if ( function_exists( 'mb_strtolower' ) ) {
                    $lowerCase = $wgLang->ucfirst( mb_strtolower( $name ) );
            } else {
                    $lowerCase = ucfirst( strtolower( $name ) );
            }

            # Make it mostly unique
            if ( $lowerCase != $friendlyName  ) {
                    $friendlyName .= '_' . substr(md5( $name ), 0, 4);
            }
            # Handle colon specially by replacing it with tilde
            # Thus we reduce the number of paths with hashes appended
            $friendlyName = str_replace( ':', '~', $friendlyName );

            return $friendlyName;
    }

How can this last str_replace "reduce the number of paths with hashes appended"? The hash append happens _before_ this line. Sorry, for my ignorance.

Sven

I think it reduces the number of paths with hashes appended compared with treating the ':' in the initial strtr. ':' is a really common character on mediawiki due to the namespace, and if it was replaced before, lots and lots of pages would enter the $lowerCase != $friendlyName check.