PHP’s mysql_connect() Reuses Connections by Default

As I mentioned yesterday, I’m doing some work in WordPress right now, and a few minutes ago I tweeted that my custom code is messing with WP’s wp_get_archives() and wp_list_categories() functions, well, I found the problem.

I am including the 4RoadService.com header & utilities files in my WordPress theme, and I am using the same user here on my test server for both the main 4RoadService.com database and the WordPress database. It turns out that when the 4RoadService.com database connection was initialized, since it uses the same connection info as the WordPress database, the existing connection was just reused, (this behaviour, by the way, is well described in the PHP documentation), then when the 4RoadService.com connection was told to use the main 4RoadService database it did, thus switching our one and only connection away from the WordPress database, and making WordPress think that there were no posts on the blog.

Fortunately, there is a quick workaround, just add one more attribute to the mysql_connect() function so it looks like this:

$dblink = mysql_connect($host, $user, $pass, true);

This way a new connection is established, and the WordPress connection is left alone.

I am left wondering why, in the loop, WordPress was able to see my posts, perhaps it establishes a second database connection in there. However, I’m not going to spend the afternoon poking through the guts of WordPress.

Endpoints: A little secret for URL manipulation in WordPress

Today I’ve been setting up WordPress as the News section of a website which loads its pages via AJAX requests whenever possible, but falls back on normal HTTP requests when the AJAX loads are not possible.

The when the AJAX requests are initiated from Javascript, /outputxml/ is added to the end of the URL. This gets translated, with some mod_rewrite magic, to a $_GET parameter called output. /outputxhtml is also possible but since that’s the default it doesn’t get used very much.

After, (mostly), building the WordPress theme I started testing, and as I expected I ran into some problems when /outputxml/ was added to the end of the WordPress URLs. I got 404 errors, which makes total sense. I thought I could get around this by simply doing a little extra mod_rewrite magic, however, it seems there’s not way to simply replace /outputxml somewhere in a url with an empty string using mod_rewrite alone. After some time, I stumbled upon an underdocumented WordPress function: WP_Rewrite::add_endpoint and its friend, add_rewrite_endpoint. These functions make it so that WordPress recognizes /category/post-name/trackback, and /category/post-name/outputxml. Excellent!

I just had to create a plugin, make sure that WordPress wouldn’t kill my $_GET[‘output’] variable, add 1 line to my .htaccess and I was good to go.

References:

And this is what my plugin looks like, (for educational purposes only. I am not distributing it):

function fourRS_outputxml_activate() {
    global $wp_rewrite;
    add_rewrite_endpoint('outputxml',array(EP_PERMALINK, EP_PAGES));
    add_rewrite_endpoint('outputxml',EP_ALL);       

    $wp_rewrite->flush_rules();
}
register_activation_hook( __FILE__, 'fourRS_outputxml_activate');


function fourRS_outputxml_deactivate() {
    global $wp_rewrite;
    $wp_rewrite->flush_rules();
}

register_deactivation_hook( __FILE__, 'fourRS_outputxml_deactivate');


/* Makes it so WP doesn't eat my nice $_GET variable */
function fourRS_parameter_queryvars( $qvars )
{
    $qvars[] = 'output';
    return $qvars;
}
add_filter('query_vars', 'fourRS_parameter_queryvars' );

Edit (August 25, 2009): Changed the attrbutes in the add_rewrite_endpoint() function.