Question

I have never used unit testing before, so I’m giving CxxTest a go. I wrote a test to check if a function correctly sorts a std::vector. First I made sure the test failed when the vector wasn’t sorted, and then as a sanity check I tested whether std::sort worked (it did, of course). So far, so good.

Then I started writing my own sorting function. However, I made a mistake and the function didn’t sort correctly. Since my test didn’t output the intermediate states of a vector as it was being sorted, it was difficult to tell where I had gone wrong in my sorting function. I ended up using cout statements (I could have used a debugger) to find my bug, and never used the unit test until after I knew my sort function worked.

Am I doing something wrong here? I thought unit testing was as simple as

1) Write test
2) Write function
3) Test function
4) If test fails, revise function
5) Repeat 3 and 4 until test passes

The process I used was more like

1) Write test
2) Write function
3) Test function
4) If test fails, debug function until it works correctly
5) Repeat 3 (even though function is already known to work)

I feel like my process was not truly TDD, because the design of my sorting function was not driven by the test I wrote. Should I have written more tests, e.g. tests that check the intermediate states of a vector as it’s being sorted?

Answer 1

Tests are not supposed to debug your code for you.

I feel like my process was not truly TDD

Your wrote a test. It found a bug. You fixed the bug. Your test passed. The system works!

This is the essence of test-driven development. Your tests tell you when you have a bug, and they tell you when you’re done.

Anyway, feeling guilt because you’re not achieving pure TDD or pure OOP or whatever is a psychological disorder. Go forth and be productive.