We went over 2,000 unit tests this past week during Iteration 72 on our Agile project. Of course, over the course of the last 18-24 months we have removed some tests, and in many cases, refactored the existing tests many times. We also have been learning a whole lot about TDD and the actual domain that we are building and testing. As we were doing this, we were implicitly discovering Test Smells, and discovering test automation patterns. The value in establishing patterns, and more precisely a pattern language in a particular domain are substantial. It's not so much that the "collector" of patterns is defining something new (some often mistakenly criticize pattern books in that regard) that you didn't know, but defining a shared terminology of our practices that we keep doing over and over. To that end, the patterns themselves not only define a shared vocabulary but serve other functions, not the least of which is learning from others. An obvious example of this is Martin's PEAA collection of patterns that enables us to say things like PageController or Lazy Load or TableDataGateway and we all know what it means. In fact, when I am talking about Interaction versus State/Behavior type of testing on CB, and others here use much of this terminology, I am in fact, talking about patterns like TestDoubles and MockObjects, among others.
When I became aware that Gerard Meszarous' xUnit Test Patterns book was going to ship Friday, I ordered it for overnight delivery on Saturday. I read well over 200 pages yesterday pretty much at one sitting, contented with a book that will change the face of the software industry, just as JUnit and all the other xUnit family have fundamentally altered software development for the better. Its definitely a big book at 944 pages, but it's not a book of excess, unnecessary pages. Rather it shows how hard it is to write defect-free software and the depth of the work that people are putting into this endeavor. The book uses Java as the language which obviously is no hardship to the C# programmer. Like most of the sound practices that have been evolving in the last ten years, this work has been evolving out of the terrific Java community.
Just like their are Code Smells, there are Test Smells, and writing good test code is just as hard and as worthy as writing good production code. Meszarous categorizes Test Smells into ProjectSmells, BehaviorSmells, and CodeSmells. Particularly interesting is his discussion in this regard to the commercial "record and playback" test automation products that have given test automation a bad name in many circles with their tendency to create FragileTests particularly with regard to Interface Sensitivity. Like many others, we were drawn in, and spent thousands of dollars with a vendor and exhibiting extreme Interface Sensitivity with the user interface. Their interface was not only unable to "pick up" most of the controls we use but even minor changes to the interface can cause tests to fail, even in circumstances in which a human user would say the test should pass. This only goes to support the notion many of us have talked about here about factoring a UI into MVP or MVC and not having logic in the "presentation."
Meszarous goes onto to provide very valuable discussions of Goals of Test Automation, Philosophies of Test Automation, and a Roadmap to Test Automation. Particularly what is important for Microsoft, and especially people like Euan who ask again about the vast difference between what we are talking about here/TDD and what they are talking about. We talk about things like Tests as Specification, also known as Executable Specification: "If we are doing test-driven development or test-first development, the tests give us a way to capture what the SUT should be doing before we start implementing it. They give us a way to specify the behavior in various scenarios captured in a form that we can then execute (essentially an "executable specification".) To ensure we are "building the right software", we must ensure that our tests reflect how the SUT will actually be used." We also talk about Tests as Documentation. Also, see Jay's piece on why even with Orcas its still not even there.
The main part of the book, of course, is the catalog of the patterns. Meszarous has provided a tremendous service to our community by not only cataloging and naming much of what we do, but also providing excellent discussions of why and how we do those things. I think, over time, this will be regarded as a seminal work in Software Development.
I'm rocking out to Do Not Pass Go by Frank Zappa from the album Guitar Disc 1