Christoph Gockel

Don't defer refactorings for too long

07 Jul 2014

The title of this blog post is a message I will say to myself for a very long time now.

The reason for this is that I hunted a bug in my Tic Tac Toe implementation for a whole day.

A. Whole. Day.

First, let me give a short introduction of the general design of my game. There are Player objects, a Board, a Game and a Rules object. While test driving the API of my Rules class I was primarily interested in whether a given board has a winner or not (i.e. rules.has_winner?(board) returning true or false).

Later then, a method like rules.winner(board) was needed to get the actual winner of the given board. In a game with just two players the method can just return two values, can’t it? It’s either the one player or the other.

Yes, but: what about a draw? The method returned nil for the case that there is no winner at all. (Note to myself: ugh!)

I felt that returning nil is probably not the best solution I could come up with, but it was “okay” for the moment.

I can refactor that at a later point in time easily” I said to myself.

As things progressed I forgot about that tiny gap I had in my Rules public API.

Remark: I like the way how Kent Beck explained it in his book “Test-Driven Development by Example”. While you are working on a particular feature, you keep notes of refactoring that you can’t make right now, but want to keep in mind.

And then it was this very gap that caused me so much trouble. Because the method didn’t differentiate between having a real winner or having a draw, the Negamax implementation chose wrong locations. But just under certain constellations of a board, it didn’t chose the wrong locations all the time. That’s why I was so confused.

After rewriting the Negamax solution several times from scratch to check that I didn’t typed or used something wrong, I just couldn’t find the problem. I even switched to a Minimax solution in between, just to verify it’s not the algorithm itself that is the problem in my code.

But after finally realising what my problem is, there was a great relief! The moment every programmer knows how it feels: You finally slayed the dragon. Made fire with your bare hands. Or, as in my case, finally fixed that not-so-clever mistake you did yourself.

I HAVE MADE FIRE (Image source: