pboyd.io

A simple case for unit tests

I learned recently that the word playwright is unrelated to write. It is, in fact, derived from wrought. A play is not written, it’s beaten into shape. In that vein, codewright ought to be a word, because I’ve never written software from top to bottom, it’s hammered and stretched until it’s right.1

My personal process usually goes like this:

  1. Figure out what the next bit of code should be
  2. Write it
  3. Run it
  4. Check the result
  5. Failed? Fix and go back to #2.
  6. Succeeded? Repeat from the top.

Yes, it’s infinite. It never ends, I just commit periodically.

Notice the tight loop between steps #3 and #5? That’s the critical path to optimize2. Running and checking the code varies a lot. For me it’s usually restarting a web server and checking the output on a few requests. But, however it looks, it takes time and it’s error-prone.

Unit tests are simply code that runs and checks your other code. Ideally, they do it quickly and completely. And they should still work the next time you need them. It’s just pre-computing what you can before that loop to avoid re-checking at every iteration. Tests won’t initiate a golden age for your company’s development department, they will bring neither fame nor fortune. In fact, sometimes they do more harm than good. But when they make sense, write them. Frankly, you don’t have time not to.


  1. In fact, can we replace all the construction metaphors with theatre metaphors? It would be a whole lot more fun. Programmers and playwrights both create scripts for others to run, so I think it works. [return]
  2. Step #1 is really where all the time goes, but it’s not so easily optimized. [return]