Using tests for feedback

Published on 20 June 2013.

Today’s thought is about using tests to give you feedback about your code.

I remember watching a screencast series with Kent Beck where he showed how he did TDD. I also remember being disappointed at the end because I felt that the tests at the last part was unfinished.

The reason I found the tests unfinished was that they depended on a server being started manually. Maybe there were other manual steps involved, I can’t remember. So what I would have liked to see in the last part was the development of some kind of setup code that enabled the tests to run fully automatically.

My first reaction when I watched the series was that the tests were not that useful since they required manual actions to be run. But what I failed to see was that the tests gave valuable feedback to Kent while he was developing his program. He wouldn’t have been able to make progress that fast and with quality without the tests.

Perhaps it was even wise to wait with the complete setup. If it turned out that the program was not useful anyway, or that the solution wasn’t good enough in some respect, it would have been a waste to write extensive setup code. When you start out, perhaps it is most important to get feedback quickly so that you can verify what you are about to do is the right thing to do. Eventually you might want full automation in your tests.

Another case where full automation is maybe not that important is when you refactor code. Perhaps the tests will only live for a short period of time, and if they can give you feedback fast enough, full automation might be a waste.

–

• Comments on Google+
• Screencast series