Read, test, don't repeat - how to avoid code complexity
Emergent Design: Lessons from Y2K
But why? What is it about bad code that makes it hard to test?
Examining this question over and over again with different customer teams has been incredibly revealing. Some of the most common reasons are:
- The class I want to test cannot be tested by itself. It has lots of dependencies that must be up-and-running for it to be able to perform its behavior. Also, this makes the tests slow to run, and therefore we can't run them very often. This makes the whole thing seem "not worth it".
- The class has so many permutations of its behavior that the test, in order to be comprehensive, must be enormous. This is usually the case when a class has multiple responsibilities, each of which can give various results; and so, there are many combinations that must be tested. A valid test must test all possible combinations because there is no way to ensure that there are no side effects inside a class.
- The issue to be tested is actually repeated in many places in the system, and so the tests will be very redundant and hard to maintain.
All of these represent design problems anyway! In the first case, I would suggest that there is too much coupling in the system if units cannot be tested in isolation (or with only a few other units in play). In the second case, I would argue that a class with multiple responsibilities is weakly cohesive in the first place, and will therefore be overly complex and poorly encapsulated. In the third case, if an issue is repeated in the system, there is a redundancy, and I know I do not want that.
In other words, considering the testability of the code I have written (or better yet, am about to write) is a great way to evaluate its quality, even if I cannot quite see how it stacks up in terms of coupling, cohesion, redundancy, and so on. If I cannot see a way to test it, maybe I should rethink what I am planning on doing.
It is like a first principle of first principles. It is one thing I can consider, which will cause me to consider other good things automatically.
Cohesion, coupling, and redundancy are like the three legs of a stool: you need to address them all if you want reliability and stability. For a while, in my courses and when consulting or mentoring, these three were the focus of my attention when it came to coding principles.
However, I have added a fourth concern recently, after having worked with teams that did not address it well, and seeing the results. Code needs to be readable.
Software developers tend to be intellectually minded people who pride themselves on the complexity of their thoughts, on being able to understand subtle, intricate, even arcane writings and discussions. It is a stimulating mental challenge to them to see how much they can do with a single line of code, or how cleverly they can accomplish a mundane task. Readability is for wimps, right? I have actually heard developers say things like: "If you can't read my code, you should not be messing with it," or "I can read my code, and that is all that matters."
The economics of software development say otherwise.
First of all, you may not be the only person who has to read your code. It is quite likely that another developer will have to maintain it in the future after you have been promoted, moved on to another project, or retired to your island in the Caribbean.
Secondly, you may have to read the code in a few months, or even in a few years, when other projects and other clever code have taken over your active memory, and at that point it can seem like reading something written by someone other than yourself.
Also, often good software is a reflection of good teamwork; your ideas may not always be the only ones, or the best ones for the part of the project you are working on. If other people cannot understand your code, how can they contribute their ideas to it? The synergy that comes from allowing multiple minds to address the same problem can be very powerful, and can lead to efficiencies that a single developer alone cannot achieve.
Next time, in our final installment, we look at the pathologies in code that indicated when redundancy, testability and readability have not been attended to.
This chapter is excerpted from the new book, Emergent Design: The Evolutionary Nature of Professional Software Development by Scott Bain, published by Addison-Wesley Professional, March 2008 ISBN 0-321-50936-6 Copyright (c) 2008 Pearson Education, Inc. For more information, please see informIT.com and Register Books.
Sponsored: Hyper-scale data management