Unit test adventures - part 1

Dieses Post stammt aus dem ursprünglichen "Mister Goodcat"-Blog bei Geeks with Blogs (2009-2010).

Some days ago I added a feature to an existing application, and when I was testing it, I felt that some calculations in the already existing parts of the software were wrong. After digging through the code I found that an external mathematical library is used, which was developed by a colleague some time ago. Due to its nature (being a pretty basic library), it was well-resourced with unit tests, and a short test run showed all of them succeeding.

I sat down and did some old-school pen and paper calculations, and the results clearly showed that the algorithm in question was way off. I went to see the colleague responsible for the library and it turned out that he found a groundbreaking new technique of creating unit tests that saved him quite some time during the development. Apparently, you just need to follow these three simple rules:

1. Write a stub for what you want to test.
2. Debug the (failing) test and...
3. Put the result of that test run as expected results into the asserts...

Very groundbreaking indeed.