Customer Reviews

Test Driven Development: By Example

5 star:		(16)
4 star:		(13)
3 star:		(4)
2 star:		(6)
1 star:		(2)

 

 

The are a small number of writers who can teach programming skills effectively. Kent Beck is one of them. There are a small set of practices that you can adopt on your own that will have an clearly observable impact on the quality of your results and the quality of your work day. Test Driven Develoment (TDD) is one of them. If you are a software developer, you want to buy, read and study this book.

TDD fits development into a three micro-phase cycle: create a test that embodies your requirement, write code that passes the test, make the code run clean. Each phase has different goals, patterns and pitfalls. Like any good coach, Beck walks you through these in detail. He uses multiple examples: most notably a business model in Java and a unit testing framework in Phython. He follows up with a question and answer section that reviews common patterns in test driven development cycle.

The level of learning involved in doing TDD is profound. The best way to read the book is to do the book. Skills come from doing not reading. I did the examples (in another language) and it made all the difference in what I learned.

A footnote for managers: TDD is the opening wedge for a set of practices known as extreme programming (XP) or agile development. Test driven development is powerful enough to work on its own for the single delevoper. If you want to realize its full value, however, you need to embrace the full set of XP practices for the whole organization.

If you've never done or are curious about TDD, this is a great book to carefully walk you through learning how and why to do it. After following its practices a bit, I've also found it an indispensible way to write new projects, modules, and code. However, the book doesn't address what happens when:
- The code base is old, and doesn't have any tests or isn't designed testable. It makes it hard to do anything other than introduce integration-level tests and tweak to success.
- You're writing UI code for a serious application. It's straightforward to solve for a dialog framework, but when you're integrating with a major windowing framework that embeds serious functionality (Avalon, in my case), there are a whole set of issues he doesn't talk about.
- Design is part of your deliverable. I don't disagree that you can get pretty reasonble designs out of TDD & refactor. But I *do* disagree that, in practice, you get designs intended to version well, that your company is willing to support for the next decade or more. I've seen the code produced, and it just doesn't happen.

A good introduction, nonetheless. But watch out before you put on the preacher-hat after reading it and doing the exercises -- at least try to do it in part of one large, real-world product.

I bought this book with high expectations. I'm a true believer in testing early and often. Basically, I think the techniques have merit, but the presentation was lacking.

I was disappointed with the writing -- all of the little asides that were attempts at humor fell flat and were distracting. The biggest disappointment was that a chapter on how to integrate these techniques into the processes of a project, especially a large project, was ommitted.

At one point, the book states rather flippantly something to the effect that "You'll have to come up with your own argument to convince your boss to let you spend the time writing all these tests." I think that since he's the one promoting these techniques, he should be able to come up with those arguments. Personally, I think the argument is something like this:
- helps produce clean, modular code by forcing the designer to think about interfaces of the objects first. Coming up with good interfaces is half the battle.
- reduces defects tremendously because automated tests are available from the beginning and are used constantly.
- combines aspects of design and testing into the construction phase. So, the argument is that the "added" time spent writing tests saves time in the design and testing phases.

I would have also liked to have seen a section on the dangers of stubbed out code. Since the technique causes you to stub out a lot of methods, as you go through the process and people are fallable, it warrants discussion. Sometimes they forget that some of the code is stubbed. I've seen situations where stubbed out code (ie, return true) provided the correct answer for a surprisingly long time. It wasn't until it got into system testing that some of the less frequently encountered data caused mysterious problems to come up in supposedly working code.

I enjoyed reading this book, however I must advise that non-coders will probably have difficulty in staying with it. I don't mean that as a put-down of any kind. It's obvious that the intended audience is the developer who is trying to understand the concept of test-driven development. A tester, however, would learn in this book that test-driven development uses tests that are different in nature and rigor than those commonly thought of as "unit tests."

I think Beck does a good job in explaining test-driven development in a way that is easy to understand. I still have some concerns about the nature of test-driven development, such as the emphasis on function over design. But I think Beck achieved a reasonable goal of presenting by example what test-driven development is all about.

The goal of test-driven development is a reasonable way to achieve "clean code that works - now." As a tester, I think the awareness of test-driven development is a good thing. I also think that this technique must be combined with other methods, such as getting quality requirements, verification and validation, to achieve a final result that meets the users' needs.

Readability - 4
Coverage of topics - 5
Depth of coverage - 5
Credibility - 5
Accuracy - 5
Relevance to software quality - 5
Overall - 5

Let me say first off that I agree with much that Kent Beck has to say: 1. Testing should be done along with the coding. 2. Use regression tests to be confident of making changes. 3. In many ways testing can be used as documentation since it is much more definitive than specification documents. 4. Testing should be used to have the client sign off on a product. In reading the book I learned the specifics of how tests are designed in TDD. It seems reasonable and I am going to make a conscious effort at designing my tests in the way suggested.

Where I disagree is in the use of the tests to drive software design. In the first part of the book, which I think is the most important part, a very good coding problem is analyzed - it is realistic, limited in scope and far from trivial. I followed along until I reached a point where things stopped making sense. I skipped ahead to see where things were headed and then things became clear.

What is being advocated is a type of bottom up design approach. This may work for some. It may even be that the book faithfully reproduced Beck's reasoning process. It does not work for me. I first have to see the larger picture, what he refers to as the "metaphor." The whole thing would have been much clearer to me if at the beginning I was told that one approach to summing money in different currencies would be to use an array to store the information but that instead the implementation would create a list similar to how things are done in LISP.

I urge the reader to judge for him/herself. Like I said this is a good example to go through. I even learned some things about more advanced uses of object oriented programming. As for software design I am going to stick with dataflow diagrams. They are still the best tool that I know of for putting together software, UML notwithstanding.

Many other reviewers have, with some justification, bemoaned the crunchingly slow pace of this book. Yes, the book moves through its examples slowly. Yes, sometimes Beck's mock humility comes off more than a little snide. It's not perfect on those counts, but please keep in mind that this is a book about a _process_, not a _result_.

The first example takes up almost half the book just to go through a pretty minimal implementation of a multi-currency representation for money. If this were a book about how to implement money representations, it would be a dismal failure. But of course, that's not the point at all -- the point is to use an example that's simple (so as not to be distracting), but just complex enough to produce adequate talking points to drive a discussion about test-driven development (TDD).

TDD is incredibly important, surprisingly late in arriving as a TLA unto itself, and Beck certainly gets points (cf. the review about "90% is just showing up") for producing a good straightforward introduction that's sorely needed. Nobody's going to come away from this book feeling filled to the brim with facts and sophisticated techniques. It's a short book (around 200 pages), and its pace is unhurried. What it does is focus on _fundamentals_.

TDD is all about buyin -- once you "get religion" and become "test-infected" (per Gamma), you've got a solid basis to grow from. It's about habits, and habits can be hard to teach. What's obvious to one person is mysterious to the next. Beck's approach of "sit here with me and listen to my thoughts on a simple, representative problem" is perfectly adequate. It concedes (repeatedly) that some of the steps are obvious, but the pages quickly and one never feels truly bogged down. He's really just teaching a handful of concepts throughout the whole book. You could write the concepts in a single paragraph; that's how much real, critical information is here. But it's _really good information_, and sometimes the key to grasping a fundamentally new (to you) viewpoint or idea is just hearing it rephrased for the 101st time, this time in words your brain is prepared to listen to.

So... it's a quick read, maybe a little pricey on that count. I'd say buy it anyway, and recoup the investment by loaning it to others on your team.

TDD is an incredibly beneficial infection; it's worth exposing yourself to a plainspoken explanation like this. You'll probably know within 50 pages whether you agree.

It's about time that someone wrote this book. Some programmers have been doing test-driven-development since the earliest days of our profession, and the rest of us have been wondering why it is so hard to development software the "traditional" (non-TDD) way.

Test-driven development (or as I prefer to call it, test-driven-design) helps you figure out the most useful interface to your class-under-test, without getting you into the psychological trap of not "really" wanting to test (and thus prove faulty) your "wonderful" code, because your code doesn't exist yet. The tests help you think about the implementation in small, mostly painless, steps.

TDD also helps you write portable code. By getting portions of the logical parts of your application done first (the "model" of "model-view-controller"), you easily keep the logic code OUT of the GUI code. Typically, programming without test-driven-design makes it too easy to put all your logic into your GUI class. Almost all books on how to use MFC and other GUI class frameworks mix the logic code with view code -- you should read this book so you can be a better programmer.

While the first two parts of the book: "The Money Example" and "The xUnit Example" may seem discontenting for an experienced XP'er, the third part: "Patterns for Test-Driven Development" is amazingly impressive. It brings lot of valuable patterns: Test-Driven Development Patterns, Red Bar Patterns, Testing Patterns, Green Bar Patterns, xUnit Patterns and Design Patterns. Despite the book "Design Patterns" seems to be provisioning, design in test-driven-development requires a slightly different look at design patterns, and Kent Beck has done his best in providing not only the common vocabulary, but a gainful technique not known to be described anywhere else before.

Before the publication of this book, there was a lack of a good manual for xUnit testing framework. The title "Testing Extreme Programming" by Lisa Crispin and Tip House, released a couple of month before this book, didn't fill the gap. This book is the first significant guidebook for xUnit ever released. While the work "Extreme Programming Installed" exposes most valuable testing experience among other XP titles, it didn't focus on xUnit as well.

I would recommend "Design Pattern" and "Refactoring" in addition to this book, assuming that you are aware of the XP manifesto: "Extreme Programming Explained".

Everyone who's read about XP has wanted a book like this forever, so a decent performance was bound to be a happy occasion (a hot dog is a feast to a starving man). I would disagree with the last reviewer and say that the first part of the book is the good part, when the author works through his 'money' example. At the end of it, there's a great moment when Beck acknowledges the importance of metaphor, claiming that he'd done the same exercise a number of times, though this time, he had picked a better metaphor and it subsequently went a lot faster. I have to laugh at this. This is a case of obiter dicta (giving away the key to something in passing). But the funny thing is that Beck doesn't notice how important it is. He proceeds to just meander through the rest of the book. Then, the book just goes down the rathole as we feel like we're being treated to a prof who's run through his material and is just waiting for the bell. The section on patterns is abominable, ending with a thing on Singleton that says something like 'Don't use global variables, your programs will be happier,' which is a ridiculous capsulization of an issue that a lot of people have discussed many times before. A Singleton is globally available. It's not a variable. Mail servers are globally available too. Guess we shouldn't use them either.

Let me also say that I am really kind of fed up with Kent Beck's Opi Taylor writing style. It's ok when the focus is on the KISS side of the equation and generally positive, but his snide little sideswipes throughout this book on everything from the open-closed principle to the idea of doing specifications (another searing, strong argument that boils down to the root of the word being the same as for the word speculation (!)) are really laughable. Makes me wanna say, yeah, who needs a spec if all they are doing is the 10 millionth payroll program or a currency converter. Don't look to Kent Beck for big answers, as a matter of fact, by his own half-conscious admission, he's as much in search of them as we are.

So why 4 stars? Per the subject, the Woody Allen principle. Another hysterical part of this book is at the very end he shows a list of things someone else suggested, none of which are covered in the book, as if he had to tell us, at the end, that he knew just how much was missing.

The book's scope is well-defined and its methodology (including a running "task list" that is updated at the end of each chapter with strikeouts and new items) is innovative. But it falls sadly short in execution.

The text is overloaded with cutesy digressions that only serve to obscure the topic at hand and irritate the reader. No "Head First" title, this. To read this book is to wish over and over that its author had had the humility to submit it to another editor's series rather than launching it under his own.

Still, there is no other book quite like it on this subject, and I can certainly recommend it for extended bookstore browsing. You may find you are less sensitive than me to Beck's assaults on clarity, in which case by all means go forth and buy.