Big, Ugly, Naggingly Red 99s

I’ve been using C# Express Edition so far to write my parser, and I’ve been running NUnit as a separate process. I haven’t been running NCover at all, so far. I like the idea of integrating NUnit and NCover into the IDE, but Express Edition doesn’t allow for plugins (as far as I know). This got me interested in SharpDevelop, which seems to have come a long way since the last time I tried it out. I got the latest version (2.1 RC 1), as well as the latest NCover (1.5.5), opened up the solution file that Express Edition had created, and everything just worked. Kudos to the SharpDevelop crew.

I ran the unit tests with code coverage from inside the IDE, and it turns out that I already have 100% code coverage. This was a pleasant surprise. There was one previous project where I used NUnit, but I only occasionally wrote tests first. In that project, reaching 80% coverage felt like a success, and I considered 90% to be a realistic goal. That was kind of silly of me; I was defining success as having confidence in only 4 out of every 5 lines of code I wrote.

On this project, reaching 100% coverage has been trivial. In fact, it’s been accidental. Since I never wrote a new line without first having a failing test, the chances of slipping in untested code would only come during the short refactoring step in each “Red, green, refactor” cycle. Since the refactoring steps are all about removing duplication, they mostly involve removing code rather than writing entirely new lines. There is always going to be a chance that a little untested code will slip through, despite taking things in these disciplined steps, but that chance is small and will bathe my screen with big, ugly, naggingly red 99s just moments after introducing the offending line of code. Over time, responding immediately to those red 99s will have a Pavlovian effect: every time I introduce untrusted code, I’ll feel the need to 1) immediately reach 100% again before proceeding and 2) crave me some Kibbles ‘n Bits.