Modify Wordpress to not replace + (plus) characters from URLs?

Question

This is happening in more than one section of our website, but in different yet similar ways.

Scenario One:

We have a URL structure like this...

mydomain.com/some-category/?subject=Some+Subject

I'm trying to get pagination properly working. I urlencode() the space and get the +, but it doesn't matter if I use a space or a + ...Wordpress always tries to do this...

mydomain.com/some-category/page/2/?subject=SomeSubject

...instead of this...

mydomain.com/some-category/page/2/?subject=Some+Subject

WordPress is removing the space/+ for some reason. We haven't figured out why just yet. Any help is greatly appreciated.

Scenario Two:

I have predefined slugs that are identical matches to our part codes in our e-commerce system. We have custom-built plugins that associate pages to the e-commerce system by way of the slug matching the part code in our system. In turn, it generates nicely formatted (and automatic) ordering tables for checkout.

Wordpress is removing any "+" that I include in the slug, which may have been part of our part codes. Is there any way to get Wordpress to quit screwing around with our URLs and let us do what we want to? We prefer to think for ourselves, rather than have Wordpress tell us how to think.

Answer 1

You can hook into WordPress’ slug algorithm and create your own slugs. For an example look into my plugin Germanix URL. I’m not sure if WordPress can accept requests with an + without further tweaking tough.

If just want to allow the slug without much coding, install Germanix and create a second plugin with this code:

<?php # -*- coding: utf-8 -*-
/*
Plugin Name: Germanix plus +
*/

add_filter( 'germanix_translit_list', 'remove_plus_from_transliteration', 10, 1);
add_filter( 'germanix_lower_ascii_regex', 'allow_plus_on_regex', 10, 1);

function remove_plus_from_transliteration( $utf8 )
{
    unset ( $utf8['+'] );
    return $utf8;
}
function allow_plus_on_regex( $regex )
{
    # $regex['pattern'] => '~([^a-z\d_+.-])~';  // was this
    $regex['pattern'] = '~([^a-z\d_+.-])~';  // now this
    return $regex;
}

remove_filter( 'editable_slug', 'urldecode' );

I haven’t tested this now but it should work. :)

Answer 2

Looks like they're fixing it in 3.3: http://core.trac.wordpress.org/ticket/18086

What they did to fix it was in /wp-includes/canonical.php, add:

$_parsed_query = array_map( 'urlencode', $_parsed_query ); 

...above this line:

$redirect_url = add_query_arg( $_parsed_query, $redirect_url );

So, yes, you're modifying core, but it's only one line and you know the change is going to stay there when you update to 3.3.