I've been making a more concerted effort to write test code as part of new modules and was doing pretty well. Then I started working on something and The Fever took hold - I just kept grinding out code. After a couple of weeks, I realized I could do a really, really big improvement with a rewrite. (Consolidating a bunch of similar things into one thing.) It's going to be so much better! Just wait and see, it really will be. But then I wasn't sure if it was really working right...and I didn't have test cases. Ouch.
So, I went back to basics. In this case, I've got a multi-process V16 bit of code that's all about message passing. So it's hard to test. I figured I'd start with 1) the most basic tests (simple permissions-like tests), 2) the easy, function-like tests (validating parameter counts, inputs, and outputs), and then 3) the hard stuff. #1 went fine, no problems. #2 the simple functions...my brain said "that stuff is fine, there's nothing to worry about there." I figured I'd do the right thing and bang out the test cases anyway. I ran them and got a bunch of bugs. I figured it was something I'd screwed up in the test cases/code itself. It took me some time to realize that, no, I'd left an enormous gaping hole in my code that wasn't visible at first sight. That lead to finding a subtle behavior in one of my validation routines that could make a bunch of other code behave unexpectedly. (I had a couple of hundred other tests cases I could run on that and did find one change I had to make. It was just a change in the error name you get for a specific sort of bad input, but still...it was a change.) So, just a good reminder of how important it is to have test code. I mean, I just found some meaningful bugs and I haven't even gotten to the hard part yet (#3.) It's so much easier to test the hard parts/integration when you've got the basics nailed down. Otherwise, you can end up chasing a bug all day only to find it was all or in part coming from something silly. Test code can be hard to write and can have it's own problems. It's also easy to not develop the tests that break your own code. Still, beats *not* having test code. ********************************************************************** 4D Internet Users Group (4D iNUG) FAQ: http://lists.4d.com/faqnug.html Archive: http://lists.4d.com/archives.html Options: http://lists.4d.com/mailman/options/4d_tech Unsub: mailto:[email protected] **********************************************************************
