byCycle.org

For a while now I’ve been planning to start using one of the many available Python Web frameworks in the development of the trip planner. Over the course of the project we have built an ad hoc, application-specific, CGI framework (of sorts). Using a “real” framework will allow us to stop refining those wheels we reinvented and instead focus more on the specifics of our application.

Deciding which framework to use has been tough. In the past year or so it seems Web frameworks have become all the rage, with several interesting Python projects and all the Ruby on Rails hype. I had initially narrowed it down to two choices, having researched just about all the competitors: TurboGears and Django. I don’t remember Pylons in that first round of research, but maybe I just missed it. Once I noticed it, it was added to the list too.

So, the idea had been floating around for a while and I had created several test applications with various frameworks but still hadn’t gotten around to making a definite decision and switching over. Recently, I listened to a podcast interview of Guido van Rossum[1]. In it he mentioned that Django is his favorite Web framework. He didn’t say much about why except that he likes the way the project is being run. (”They really get open source.”)

Having heard that and having also narrowed the field to just Pylons versus Django, I decided, “OK, well I’m going to go ahead with Django then.” I ordered up a new Ubuntu VPS[2] to host the new version of the trip planner on and got to work. It was exciting to finally make a decision and move forward.

Getting started was easy and I really liked the admin interface, but about five minutes into it, I realized Django wasn’t going to work for me. There was only one reason: the tightly coupled ORM. Sure, I could just not use that part and use a different ORM, but somehow that seems “impure.” Maybe that seems silly, but remember that I had only decided to use Django on, essentially, a whim.

The reason I don’t want to use the built-in ORM is, I want the Web interface to be orthogonal to the core functionality. I want them completely separate, as they are now in the CGI version. I want to keep database access in the back end and be able to create other interfaces as necessary. Now, I suppose I could just use the Django ORM in the back end (?), but there’s another ORM you may have heard of, SQLAlchemy, that I was/am interested in trying out.

This led me back to Pylons and thinking about some of its good points: it’s WSGI from the start, Routes seems like a really cool way to map URLs, and the (Myghty) templates use Python.

Thinking about it, Pylons seems more like the “natural” choice for this project and its architecture. On the other hand, I think Django would be great for projects that don’t need such separation between the front and back ends (or where there is no back end as such).

I’ve just gotten started with the CGI=>Pylons conversion and so far it’s been a pleasure. I’ll report in more detail once I’ve gotten further into it. One “gotcha” that came up, that I don’t think this is specific to Pylons, was creating a controller with the same name as the top-level package. Don’t do something like this:

paster create –template=pylons tripplanner
paster controller tripplanner

Python will be unable to import the controller.

[1] The creator of Python (which I mention for non-technical readers)
[2] vpslink.com (seems good so far)