Good Habits, Techniques and Workflows (GDD#2)

By | 2013-02-10

Welcome to another article in the Game Development Design series. This article covers good habits, techniques and workflows which are important for a general programming behavior and how to approach developing games (or even applications in general) in a way that will improve your progress.

We will talk about some well established techniques in the field of rapid development, which will move your programming focus to implementing features that are really required, and that those implementations are guaranteed to work as best as possible by automatically checking them.

In my opinion, the typical game developer who makes games as a hobby/in an indie fashion works like this: You have a very rough idea, you think it might be funny/interesting, you create the project, you start programming on the core features (resource management, state managers, basic rendering etc.), then move on adding features which you think might be useful, and possibly implement the first game logic snippets.

During development, the code is tested by just running it and manually validating its functioning. For example one feature request is rendering a tilemap. If you see the tilemap on screen, you’re done. Sometimes you might also add some features just because you think they are interesting to make.

Personally I’m convinced that such approaches are really not good. Instead they promote losing focus of what you are really targeting for — making a game.

YAGNI

The first principle I’d like to mention is YAGNI, which means “You Ain’t Gonna Need It”. Before implementing a feature, you should make sure that it’s actually needed. A game taking place in caves does not need skyboxes. You might decide to add some outdoor scenes later in the development process, but for now, you don’t need them.

A very similar idiom is “Make games, not engines”. Unfortunately a lot of people are indeed writing engines, and totally forget about the fact that they actually want to make a game.

So, here’s what you do: Break down your game idea into smaller milestones, but don’t think too much about things that are too far in the future or even undecided. A first milestone for a cave game could contain rendering a cave and moving around in it. This requires handling basic input, moving a camera, and rendering a cave with a simple texture (or even a solid color!).

When you implement the mentioned features, you will exactly see the requirements of the engine. In this case you need a renderer with texture management, for example, so you add it. But you don’t need shader support yet! Even if you’re sure that you will need it later, you still don’t need it now, so drop it!

Working the YAGNI-way ensures that you focus on small things; you keep the scope tight. Our brains tend to think big and try to put all possible requirements into one solution. Unfortunately this will never ever work out. So give it something smaller to focus on.

Refactoring

Some of you might already be sighing, some of you not. If you are sighing and know what refactoring is, I’m very sure you don’t do it (and by doing it I mean regularly, not for parts that are completely messed up and don’t leave you an option except refactoring).

Refactoring is the process of taking old parts of code and moving, rewriting or removing them. Often refactoring means a fair amount of work. Work on something that’s usually already done, which makes it boring for a lot of people, and thus they’re avoiding it whenever they can. For example if you’re refactoring the implementation of a specific feature, you will most likely not change the feature itself, but just the code that implements it. This can range from renaming a class to completely rewriting it.

However: Refactoring is important and it has to be done. In short intervals, regularly, by every developer, in every project, always, without exceptions.

In my opinion there are two main reasons for that:

  1. Projects are mostly not finished within days. Instead they often need weeks, months or even years to hit the release day. During that time, every single developer who’s taking part will improve in programming. I’m very sure you know the feeling when you look at old code of yourself (like 1-2 years old), screaming “What the heck did I do there?!”. The good thing is that it’s mostly code of dead projects. The bad thing is: It will of course happen in active projects too! So make sure that as soon as you think something’s “strange” or “not good”, you will change it.
  2. Source code is never finished, especially when working in teams. Others might add code to classes you write, or you extend them yourself. It’s very common for people to think “Oh well, I just need this one additional function, so I’ll just add it to this class, because it will not require much time, and the class already exists”. This should not happen, but it will, and you’re sometimes even not aware of doing it. Whenever you think something like “This doesn’t look right, but it works, so I’ll use it”, stop your work and refactor.

To sum it up, refactoring results in cleaner code. You should not allow dead and unmaintained code. You don’t want ugly code, you want to do your programming job right. No excuses here, refactoring needs to be done.

As a bonus to cleaner code you will also get a lot of success moments. Everytime you finish refactoring parts of your code, you will feel that you just removed something that was not right. It will make working with your codebase a lot more fun, because you’ll feel that everything is in a good shape. And if it’s not…you will know what to do.

Testing

Again I can hear some of you sighing, because I will now mention the evil U word: Unit tests. To be honest it’s a shame that a high percentage of all developers are not testing their software except running the program/game and validating the results manually.

Unit tests are programs that test your code for expected output. For example a function add( a, b ) is supposed to sum two integers. A unit test will make sure that add( 4, 5 ) returns 9.

There’s only one rule for unit tests: You have to test every single line of your code. This basically means that your tests have to make sure that all lines of all functions are triggered. This is also called “test coverage”, and your goal is to get as close to 100% test coverage as possible.

I can already hear the excuses of people who know unit testing but don’t do it:

  • It’s time-consuming and boring.
  • It’s not needed for small projects.
  • Unit tests themselves can contain bugs, so tests are not safe.

Yes, it is time-consuming, and for people just getting into it, it might also be boring. However, spending time on something that contributes in making your software more bug-free is good invested time. You want to be a good programmer and you want to care about that your code does what it’s supposed to do, in every possible situation.

It doesn’t matter if you’re writing a small project or a big one. There’s really no good reason to allow a higher chance of bugs and crashes just because a project is small.

Unit tests are written by humans. That implies that unit tests themselves might contain bugs. However unit tests are not unit-tested — if you’d do that, you will also need unit tests for unit tests, and unit tests for unit tests for unit tests. That’s not practical. The simple rule to avoid bugs in tests is: Pay attention when writing them. Anyway, what’s better: Not testing your software at all or accepting the rare case of having a bug in a unit test?

Adding unit testing to your projects gives you a lot of benefits. At first you can be relatively sure that all your implementations work as supposed. Secondly unit tests avoid being hit by side effects. Software can and usually will get very complex. It’s impossible for developers to ensure at any time that the whole code is sane and working. Unit tests however are a great tool for detecting such errors.

Let’s say you wrote a function at the beginning of your project, and 3 months later you change it. You also adjust the unit test that validates that single function. But of course a lot of other functions might use it as well. If any other function will not work as before, you will immediately see that by failing tests. This is truly minimizing side effects and typical “Oh, how could that happen?” moments.

Another huge advantage is the feeling you get by testing your code: If your tests have a good test coverage, you won’t get that typical “Hopefully my changes do not break anything” thoughts when adding new code to your project. Instead you will just feel good, because you saw the “All tests passed.” message before. This is also related to the “Our brain wants to cover everything but can’t” problem mentioned in the refactoring part before: When not testing, the brain will often tell you that your changes might contain or provoke bugs somewhere else, which results in an uncomfortable feeling. Giving that responsibility to automated tests will reduce your stress level in that regard a lot!

There exist a lot of C++ solutions for doing unit testing. Really simple but not so comfortable solutions are using C++ features like “assert()”. However I’d really recommend using a specialized library for testing. Personally I like Boost’s unit testing framework. What you use is completely up to you, as long as you test.

Test-driven development

Let’s take the unit testing one step further: Test-driven development is a very established programming workflow, which is getting more and more popular these days, especially for rapid-development-oriented teams/developers. The required steps of the workflow are very easy:

  1. Write a wrong test case before you implement anything.
  2. Implement the feature for which your wrote the test.
  3. Run the tests and see them fail (due to wrong written test case).
  4. Fix test case.
  5. Run the tests again. If they still fail, validate that your test case is correct. If it is, fix the implementation.
  6. Repeat 5. until all tests pass.

Why do you have to write a wrong test case? Because you want to make sure that the test case is actually triggered. It might sound funny, but things like forgetting to add the test case’s .cpp file to the build system do happen from time to time.

So, what are the advantages of test-driven development?

It ensures a very high and close to 100% test coverage. If you only implement things for which unit tests exist, they will always be covered by tests.

You automatically write examples and documentation in form of source code. This is very useful for other developers on your team and even yourself. Instead of looking at API documentation only, developers can open the proper test case and see how it uses the production code.

By at first writing unit tests before the implementations, you force yourself to think about how your code shall look like, i.e. how you want to use your new feature. This highly reduces the risk of having to refactor stuff later. It also contributes to YAGNI, because you often find yourself writing test cases for things you don’t really need, and that’s really easier when writing a use case (the test in this case) instead of the implementation only.

Your functions will be better testable, because they just have to. People who are testing, but not doing test-driven development often have hard times testing their features, because they are not atomic/isolated enough. When writing test cases before implementing, you will automatically try to make them as simple as possible, thus resulting in very small-scoped functions.

Last but not least, it’s fun! What’s “time-consuming” and probably “boring” in the beginning will evolve to something fundamental. With test-driven development, you’re usually spending more time on writing use cases and validating that your code works. When you’re finished, and all your tests pass, you can just go ahead and use your sane code in production. Again a good feeling!

Conclusion

This article is about well-established techniques and habits for programmers, where a lot of them are taken from the rapid development field. It gives suggestions for writing cleaner code, making code clean again and making sure that code works.

Tons of other habits and techniques exist, and it’s really impossible to cover them all. But as long as you try out and adapt some of them (and please, don’t give up after a few days only because you’re used to old habits), your code will be much cleaner and better maintainable in the (near) future.

If you like the GDD articles and would like to support the author, consider buying the PDF/EPUB/MOBI version in a pay-what-you-want fashion:

Leave a Reply

Your email address will not be published. Required fields are marked *