The Twilight Report

One of the more decisive things I did early on working at s-mart was to write a fairly comprehensive test suite for the provisioning code for which I am responsible. Before that we didn't have any regression tests at all, so it wasn't something that I had to do, but it seemed like a good thing to do. I'd say that it's been a resounding success, because although it took me a fair amount of time to write the test cases, it's saved me lots of time in the long run. I can make changes to the base classes pretty easily, run the entire test suite and be fairly comfortable that I haven't broken anything. Not even "anything major" but "anything".

The reason I thought this was important was from my work at The Company, where we had pretty good regression tests, and my work at Company 2 where they didn't have any regression tests, and their approach to not breaking existing systems was to simply fork the source code any time they needed to make a customization. At The Company we even had nightly regressions, which meant in the morning you could see what you broke the night before. I'd like to set up something similar at s-mart, but I haven't had the free time to do it. It certainly wouldn't hurt, the regressions take a long time on the lumbering old development machine, but nightly regressions would run at some ungodly hour in the morning when nobody is around anyway.

Today, though, I felt vindicated in my approach when Andrew got Newt to write some test cases for the code that he's been working on, without any prompting from me. Even though Andrew doesn't like writing test cases, he sees the value in them, from the way that I've used them. It will only be a total victory when I get Andrew himself to contribute to the test suite, but I'll take what I can get :)