This is going to be a short one, but I think the changes to GitStore are cool enough to deserve a little blog-squeal.
So, yeah, beginning to play with
GitStore for Project Behemorshmallow, I realized that the module had a pretty serious monkeywrench to throw in my delicate machinery. It turns out that it was playing fast and loose with file definitions within the git repository. Instead of having, say, a file named foo in a directory bar, it was storing the file named bar/foo in the root directory. It’s no problem if you just use
GitStore to get at your data but more of a uh-oh if you are also manually editing that repository.
At the same time, I also realized that upon creating the store, all its stored objects were deserialized and kept on the ready in the store object. It’s not problematic for small repositories, but obviously that wouldn’t scale very well.
Sooo… I went to town on
GitStore yesterday and today, in-between an Alien marathon, a wedding anniversary, and a soccer match. A new version of the distribution is now on its way to CPAN, solving all those pesky problems.
One point that should be mentioned is that the code churn of the new release touches close to 50% of main module’s lines. And yet, I’m not too overly worried. Why? Because the distribution has an awesome testsuite that covers 86% of the codebase (100% of the functions are at least visited), covering all the basic scenarios. All those tests not only fill me with cool, smooth confidence, but they also greatly simplified the refactoring process. Instead of having to think of all the repercussions some core changes would bring (well, maybe not totally instead, but rather in addition to), I had the luxury to make changes, run the tests, see which corners of the codebase went boom, bring my attention to them, fix them, and repeat the process all over again — basically, it cut the larger problem into ones that are smaller and easier to handle.
I guess that what I’m trying to say here is: automated testing? It’s the bee’s knees.
Leave a Reply