bluefeet Yet another Perl Hacker weblog

Scientific Troubleshooting

Over the years I’ve naturally applied scientific principles to the way I program. There are two main principles that I think most good programmers ought to know.

Occam’s Razor (Wikipedia)
The idea behind Occam’s Razor, when applied to programming, is that when you are presented with a problem you should pick the simplest solution first and go try and prove it. We naturally formulate possible reasons (hypothesis) in our mind as soon as we are presented with a problem. Most junior programmers will pick either the first, or the easiest, hypothesus and go forward blindly. Neither of these is the simplest.

The easiest one is not a good choice because easy has a different meaning for every programmer in every different situation. Easy could mean that you are willing to sacrifice hours injecting print statements all over 20,000 lines of code. Easy could mean that you don’t understand the underlying system, so, rather than dig in a little and understand how things are working behind the scenes, you opt to stay safe and kick the tires hoping that something will present itself to you.

Picking the first hypothesis is an even worse choice because you’re not even attempting to prioritize the steps you are going to take to find a solution for your problem. Instead you are just blindly jumping in and will likely waste hours determining that your first hypothesis was not correct.

The bottom line is when you are presented with a problem it is important to consider several possible reasons for the problem, choose the simplest reason, and then go forward. More often than not the simplest solution is also the fastest one to confirm, which allows you to get through your initial list of hypothesis fairly quickly. At this point you will have identified the problem, or will have ruled out a bunch of stuff that will allow you to better test your larger hypothesis.

Scientific Method (Wikipedia)
The scientific method, when applied to programming, is a set of steps to take when troubleshooting an issue. In many cases following the scientific method is overkill, as abiding by Occam’s Razor is usually more than enough. For larger issue it helps to apply the scientific method which helps you structure your course of action in to various steps and can give you a visual map of all the steps you need to take to diagnose, and fix, your issue. The steps are:

1. Purpose – State your issue.
2. Research – Find out about the topic, if necessary.
3. Hypothesis – Predict a reason for the issue.
4. Experiment – Develop a procedure to test the Hypothesis.
5. Analysis – Record the results of the experiment.
6. Conclusion – Compare the hypothesis to the experiment’s conclusion.

It is important to come up with several hypothesis, ranging from very simple to very complex, and repeat steps 3 through 6 for each one. Also, remember Occam’s Razor when picking which Hypothesis to test first.

I find it works best for me when I get a pen and a pad of paper. Steps 1 and 2 will be documented on the first piece of paper. Then each hypothesis, steps 3 through 6, will get their own sheet of paper. Sometimes I can get everything to fit on one sheet, but I usually have to overflow on to a second piece of paper anyways, so its good to plan for that up front.

These techniques apply to any programming language such as Perl, JavaScript, HTML, CSS, SQL, etc, and have saved me, and can save you, countless hours and frustration.