Trying out Test-Driven Development

Question

After reading this post I kinda felt in the same position as the guy who asked the question. I love technology and coming up with new ideas to solve real world problems just gets my neurons horny, but the other part of the equation - actually getting things done (fast) - is normally a pain in the ass to accomplish, specially when I'm doing this for myself.

Sometimes I kinda feel plain bored with code, some other times I spend more time moving the cursor in the text editor and staring at my code, trying to come up with a solution that is better than the one I already have. I heard this is a disease called perfectionism.

I've read in that same post (and also a few times here on SO too) that TDD is actually good to stop coding like a girl, however I've never given a chance at TDD - either because I'm too lazy to learn / set it up or because I don't think I need it because I can do all the tests I need inside my head.

• Do you also believe that TDD actually helps to GTD?
• What do I need to know about TDD?
• What about alternatives to TDD?
• What would be the best methodology to organize / develop a TDD web app?
• What libraries should I use (if any) to make my life easier?

PS: I'm primarily (but not exclusively) working with PHP here.

Answer 1

Personally I think TDD is at best overkill and at worst an impediment to a the creative process of programming. Time that is spent laboriously writing unit tests for each as yet unwritten methods/classes would be better spent solving the original problem. That being said I am a big fan of unit tests and believe wholeheartedly in them. If I have a particularly complex or troublesome piece of code I'm more than happy to write 20 unit tests for a single method but generally AFTER I have solved the problem. TDD, just like every other programming paradigm, is no silver bullet. If is suits you use it if not keep looking.

But take my opinion with a grain of salt. A much more interesting one comes from Kent Beck and How deep are your unit tests?.

Answer 2

Do you also believe that TDD actually helps to GTD? My biggest concern was simply not being able to test code. It was too complex. Our core libraries weren't built around an easily testable interface. So we wrote tests for whatever we could. In the end we ended up refactoring our core libraries to make life easier. Besides that, it's a change of a mindset, and i would definitely consider allocating more time on your first TDD project just to kind of flush out some of the problems you may have along the way.

What do I need to know about TDD? TDD is not a substitute for a methodology. It is a beneficial addition or at least it's supposed to be. If done right, TDD greatly improves software design. It also acts as your internal documentation. If you want somebody to look at your API and understand how it works they can simply look at your well named an formed tests.

What about alternatives to TDD? Like i said, i wouldn't consider this a substitute for a methodology. There is an alternative and that is not to use it :)

What would be the best methodology to organize / develop a TDD web app? We have been fairly successful with scrum/agile, if that's what you're asking.

What libraries should I use (if any) to make my life easier? My PHP knowledge has expired 5 years ago, and i'll let somebody else answer this.

Either way, just my 2 cents. If you feel like reading here is a good overview: http://www.agiledata.org/essays/tdd.html

Answer 3

I've recently started using "fat models thin controllers" http://www.amitshah.in/php/controller-vs-model.html : shovelling as much code into the model (and out of the view / controller) as possible.

I use PHPUnit (and the Zend Framework support for it) to test only a few complex models in my web apps. Writing unit tests that exhaustively check a 2 line function that executes a simple SQL query is a waste of time IMHO. Over the last couple of years I've got lazier and lazier with writing tests, and for most web apps its not worth it because the code is so simple.

But in a recent project (an ecommerce site that tracks stock levels over multiple warehouses with composite products) I started doing test-first development because I knew there was going to be a lot of subtle complexity that I couldn't keep in my head all at once. The only way to keep everything working as my thinking developed was to write some tests. With some parts it seemed more natural to write the class then the tests, other parts were test first, yet other parts didn't need tests because they were trivial. TDD is a tool, not a religion. Use it when it works, stop if it doesn't.

Where I see the benefit in TDD is in reducing complexity and increasing speed (the rate at which I can solve problems). If I write some tests that prove my code is working, then as soon as the all the tests pass, I can move on to the next problem. This puts the fun back in coding for me.