Tagadelic: TDD, OOP and seeking maintainer

TL;DR: Tagadelic is close to a Drupal 7 release, with an easy upgrade path to Drupal 8. It is completely rewritten, Object Oriented and Test Driven. Since I don't do any Drupal anymore, I am looking for someone who can maintain a clean, OOP and TDD-module, to take it over from me.

There is just so much you can do when porting an age-old module again and again. Tagadelic has been around since mid-2005, has been ported over and over again. Mind you: not upgraded but ported. Quick, dirty and "works-for-me" ports. Like most other modules, actually. There never was a stable release for Drupal7, because the module never was really stable in the first place. Yes, it might work (for you), but that is far from stable and releasable.

Between 2005 and now, I learned programming properly. I mean, OOP, unit-testing, patterns and all that (This also lead to me, abandoning all my Drupal work, mostly).

None of my publicly released Drupal-modules ever resembled that progress I made; mainly because Drupal itself is not OOP. Has poor testing abilities (please read on, I will explain later) and applies quite a few anti-patterns. This makes writing really clean and pretty code, somewhat discouraging. Most of its examples, best practices and defaults go straight against what is in general considered best practice.

But since Tagadelic is used by a lot of people, I wanted to create a proper replacement. A module with pretty code, easy to implement APIs and some additional, turnkey modules for those who cannot or will not write these few lines of PHP. A module that resembles what I now consider good code and properly developed. As a replacement for what I thought proper 8 years ago.

I coded for several months and today released the first alpha.

In the long run, I can conclude three things:

Drupal is not really ready for OOP development. The interfacing between my module and Drupal required me to write wrappers (so that Tagadelic classes access Drupal-functions in an OOP-manner) and to write the modules themselves with global functions, since that is how Drupal expects the hooks and implementations.

DrupalWebTest is way, way too slow and feature-poor for Test Driven Development (TDD). Tagadelic only has about 150 DrupalWebTests, but running them all takes over 5 minutes (on my machine: quad-core Intel 2.67GHz, SSD drives only). Note that in a typical Rails (being -rightfully- known for being very slow) with cucumber suite of over 600 tests takes under 30 seconds; that includes selenium opening Firefox and clicking around in a few tests. 30 seconds is considered unacceptably slow there.

When developing test-driven (or Behaviour Driven) you typically run the isolated tests five, six times. And the entire suite of tests at least once. So aside from the actual coding, the testing alone takes 30 minutes. This is both discouraging (meh, I'll just assume everything is still green, will test in next iteration) and very hard for your "flow" and concentration. It is feature-poor in a sense that I ended up writing most assertions and several set-up functions myself. assertXpath()? Nope. assertHasId()? Nope. assertIdenticalArrays()? Nope. And worse is that it breaks a very important rule for testing: isolation. If you want to test whether some admin-setting can be saved and creates the proper variable, you are also testing whether a Drupal is installed properly, user can log in, is admin, can access a page, has nodes, has access to creating these nodes and so on. I ended up poking into the database (not even "my" tables) because somewhere in the clutter of setup-tasks stuff was created but it failed.

It is really fun to write unit-tests with phpUnit. I was very much positively impressed by that test-environment and by using it. The biggest adventure was how to stub out Drupal. Drupal, using global functions for stuff like check_plain() is nearly impossible to mock and stub. I solved this by extending my DrupalWrapper and stubbing that. After all: I don't care whether check_plain() itself works and clears out XSS, I only care whether or not my classes call that function in proper places to ensure clean output. Testing whether check_plain() works is not my concern, here. I chose phpUnit over DrupalUnit, because the latter is pretty much unusable for unit-testing of arbitrary classes.

And now it is time for someone, or several someones to slowly take over the module. Together we will release a Drupal7 2.0 version and then I can carry over all project rights on my last Drupalproject.

Interested? You should be:

  • Familiar with PHP OOP development. You should probably feel that the usual way of Developing Drupal modules in a none-OOP manner is not a very good way.
  • Familiar with PHPunit and Drupal Tests. You should feel strongly for TDD and good test-coverage. You should probably feel that even though writing Drupal Web Tests is not (yet) perfect and requires time and effort, it always should happen.
  • Able to maintain such a module for a substantial time. It being TDD and all, means that it won't take you a lot of effort or time. But it would be a shame if three months after a release you abandon it altogether because you like Node.js better. Or so.
  • Wanting to develop on Github. At least until the 2.0 release.
Woodcut from Doré. Purely illustrative
Doré Woodcut. Its only function is to make the layout look better. And these images are really nice themselves

About the author: Bèr Kessels is an experienced webdeveloper with a great passion for technology and Open Source. A golden combination to implement that technology in a good and efficient way. Follow @berkes on Twitter. Or read more about Bèr.

blog comments powered by Disqus