To Evolve Your Code, Let Names Change

Abstract

The names of things in code reflect what those things do. For code to evolve, the things in it must change; often, those changed things ought to change their names to reflect what they have come to do. If you do not promote name changes when needed, then natural development will become stifled.

What then, can you do to promote name changes? Here are three things:
1. Give things abstract names. Call the thing Alpha. When you know better what it will do, you can give it a better name.
2. Write tests; make commits. If you do this then you can make changes at less cost and risk.
3. Make efforts to get good at using tools for renaming. You may tend to avoid this because (a) you have to learn usage and (b) you fear making a mistake could be costly. But if you have done #2 then you have nothing to fear.

If you make a name change when you feel it is appropriate, it may cost you some time and effort in the short run, but in the long run your development will have better flow.

Exposition

Names carry information. We strive for readable code. As far as names go, readability means the role or purpose of a thing is reflected by its name. Say again, the name of a thing informs us as to what it does and what it is for. This information is conveyed by the name to the person reading the code.

Now, if you knew what everything was to be ahead of time then you could devise names at the beginning and never have to change them. But that is unrealistic. Code has to be refactored.

Imagine that your code is a number. Say it is 130. That number has some factors. The names of the factors are: 2, 5, 13. Now imagine that your code wants to naturally evolve to another number. Say that is 140. The factors are different. If you insist on keeping 13 as a factor you will never get to 7 and you will never get your code where it wants to get to.

Now think about your factors as software components with names that reflect what they do. If you limit your refactoring to only be aggregation or adding new components, if you prohibit changing what things do because you are unwilling to change the name, then you may well be restricting the natural development of your code.

Why do that? Because we act as if we know what are doing in order to be able to form a closure in our minds and get going with doing something. The problem comes from taking that too far. The reality is that we learn as we work; new information is accounted for by changes to the code. Changes do have costs but some change will happen anyways, even if you do not want to have to pay.

Open

  • ? Besides names, what else can stand to be changed?


« index »

« subscribe »