Archive for March, 2011

Side-stepping the absolute URL requirement in WordPress

This is a limited use-case, but it’s one I’ve just successfully figured out how to deal with. Hope it helps.

I have a WordPress codebase that is shared between three domains. The code executes the same on all three domains — only the domain itself is different. Which, there should be nothing inherently problematic about doing this.

Problem is, when you set up your blog’s settings, WordPress requires that you enter a fully-qualified absolute URL for the blog’s root directory — and that URL is inserted throughout every nook and cranny of the app. For whatever reason, you can’t supply it a domain-less path, even though it seems like this would be a perfectly reasonable thing to expect. (I read one explanation that this would be too problematic for many users.) There are way too many references to this absolute URL to fix by hand, and if it can be avoided, it’s not generally a good idea to go monkeying around in the guts of the app to try to change its behavior. Maintenance becomes a lot more hands-on. Upgrading can easily become a migraine.

So: database hacks to the rescue.

You can actually force WordPress to use a domain-less root path, by altering two settings directly in the database. In the wp_options table, change the siteurl and home values to whatever URL you like. Like, say, the perfectly reasonable / or /blog. The only downside is that if you ever try to change anything on the General Settings page from within wp-admin, WordPress will bark at you and give you a couple of error messages. It’s okay. You can ignore them. Everything else works just fine.

Update: Apparently, everything else does not work just fine. Setting the URL in the database broke Akismet.

Akismet requires the home value in the database options to be the URL associated with your API key. Unsurprisingly, if it doesn’t have that URL, the service requests will fail.

WordPress is great and all, but like every other off-the-shelf software package, there are times where if you want to customize it and color outside the lines they’ve drawn for you, you end up diving down a frustrating rabbit hole.

I’m sure I’ll port this bad boy over to Django soon enough. Then: sweet, satisfying, complete control.

Integrating Jinja2 and Django

UPDATE: Here’s a smart drop-in app that integrates Jinja2 and Django, with a minimum of fuss.  By default, it inspects the filenames of your templates to determine which rendering engine to use.  So if you want to use Jinja for a particular template, name it “foo.jinja.html”, for example. Otherwise, you just leave the template as-is, and the default Django rendering engine will do its thing.

Although I haven’t explored this feature, the author suggests you can use any other logic you’d like to determine when to enable Jinja rendering, such as a tag at the top of the file, for example.  Me, I’m satisfied with the filename extension, because, hey, that’s about as simple as it gets.  Pretty, pretty clever.

 

Jinja2 is a real improvement over the standard Django template system. It has a number of useful features that the built-in Django templates just don’t have — especially when it comes to basic visual markup logic (elif statements, basic macros). Thankfully, Django is built to be pluggable and modular, and it’s not terribly difficult to get Jinja2 working with it.

When I wrote my first Jinja2-based app a couple of years ago, I did so using a little helper script I cobbled together from various Jinja2/Django-type pioneers, and the result was a workable — if a little hacky and kludgy — module that I could reuse in each of my apps. I import it in my views.py, and it gives me the most basic template-type methods like render_to_response and such. It’s limited, but I really haven’t had need for much else. (And when I’ve needed to expand its functionality, I have.) But it’s always felt like a temporary fix — a little incomplete and hackish. I took a look around recently to see if there were any other solutions that integrated Jinja2 into Django a little more gracefully and completely.

It wasn’t long before I found coffin — a library that’s more or less a port of Django’s template functionality, but using Jinja2’s rendering. Its real killer feature, IMO, is how easy it is to use Django template tags and filters within coffin — which is something you might not expect to need, but it’s very handy to have. For example, coffin includes a ton of useful tags and filters out of the box — like the super-useful {% spaceless %}. And you can easily add others, without having to write a complete port.

Coffin works a lot like my little helper module. You import it in every page, explicitly.

But I found something that I think is just a tad better: this fork of coffin. Its best feature is that it can be plugged in directly to Django’s native template loading system, so that it is integrated a little more neatly. All it takes is just a little adjustment to your project’s settings file:


TEMPLATE_LOADERS = (
'coffin.template.loaders.Loader',
)

JINJA2_TEMPLATE_LOADERS = (
‘django.template.loaders.app_directories.Loader’,
‘django.template.loaders.filesystem.Loader’,
)

The coffin template handling is enabled more or less transparently — using Django’s syntax while the templates are handled behind the scenes. Thus, in your apps, instead of this:

from coffin.shortcuts import render_to_response

You can use the usual Django syntax:

from django.shortcuts import render_to_response

Helpfully, this fork also allows you to turn off Jinja2 rendering on an app-by-app basis, with a simple tuple in your settings file. So it’s easy to have, say, the Django debug toolbar running with Django templates alongside your app that uses Jinja2 templates.

I really like this approach, mainly for the reason that enabling Jinja2 can be enabled — and overridden — completely from the settings file.

I don’t have a strong opinion about the pros and cons of these two approaches, but I’ll be trying out the template-loader fork of coffin to see how well it runs and how easy it is to use. I’ll be curious to see if anyone has any input as to whether the template-loader approach has any design concerns that I haven’t considered.

Meanwhile, give it a shot! Enjoy.

Return top