Spent a lot of this week fixing several bugs in csvbase that people found.
I've always done test-first development. csvbase apparently has 89% test coverage (I just checked) and most of that 11% is non-production stuff or experiments. The test base has caught a huge number of bugs and regressions but still - a lot of mistakes get through.
I've never understood how anyone manages to write real production code without tests. Am I doing something wrong?
