This week's book giveaway is in the OO, Patterns, UML and Refactoring forum. We're giving away four copies of Refactoring for Software Design Smells: Managing Technical Debt and have Girish Suryanarayana, Ganesh Samarthyam & Tushar Sharma on-line! See this thread for details.
Marilyn , unit test is test code. As per XP principles , you will first have to write Test code for your sotware , before you actually start coding. You have to continuously to the testing the code you write & make changes there itself. It is not like your usual watefall software lifecycle , where you do testing after the prodcu is done. all programming is done. i hope this is helpful. your bartender. SHAILESH
I think concepts like this are often best explained using examples. Let's take a very simple piece of software develeopment :- a Java class to parse a line of comma-separated text into an array of Strings. If I assume an XP-style process, the first thing to do is to produce unit tests. This may seem wierd, but bear with me. I'll write the unit tests as fragments of code which could be used by JUnit (a unit-test framework used by many XP/Java programmers). OK, so what would be the simplest unit test to start with? How about checking that an empty line generates an empty array?
This shows several very important things about XP unit tests. 1. Unit tests are just code. For XP to achieve the full speed and quality of code, you need to be able to run the whole unit test suite quickly and often. So use a unit-test framework which lets you write executable tests. 2. Writing unit tests as code has already made me think about what interface my little thing should provide, and what to call it. Even though neither the source code or the class file exists yet, it has a name, a constructor and a method. Design has already started! 3. Always start as simple as possible. I've not worried at all about where the class comes from, how it's loaded, what other methods it provides, what interfaces it implements, what happens if I give it any other input... I'm only testing one thing - how it behaves when given an empty string. 4. A single "unit test" can be several "asserts". In this case, in order to check that the length is 0, I also check that the returned array is not null. 5. This test will fail. In fact it won't even compile! But that's OK. It's what I expected when I wrote it. So now, having written the test, I can proceed to the next stage of this little task; implementation. XP recommends wrinting the simplest possible implementation which passes all the tests so far, so let's write a simple class to do just that.
If I compile the code and the test case, then run it I should get an indication that 1 of 1 tests pass. Excellent, the first iteration is complete. If I get any errors, either at compile time, or run time, I fix the code until all tests pass, then move on. Now what's the next unit test? How about "if the line has only one column, return it"? Off we go again; add another unit test:
This looks good, but if you look closer, the principle of Once and Only Once has been violated. A new LineParser is created in both test cases. So let's refactor a little. Luckily, JUnit provides a place to put common initialization code; the method setUp(), so our whole test suite now looks like:
Now we get to update the code to implement the new functionality. because the original unit test is still in the test suite, it will make sure that our new change still works for empty strings. So here's the new code:
I guess you are getting worried by now. Most experienced programmers are already trying to solve the whole problem, but this is the very heart of XP. We don't know what the next unit test will be, so we should not attempt to solve it. XP is based on being able to deliver each new functionality as soon as possible, and allowing the customer to cancel development at any stage and always have a working version which implements all the features so far. So the extra 20-30 minutes it might take to sort out how to use a StringTokenizer, how to decide how many array slots to create and so on would be wasted time, if the customer only wants files with no words or one word per line, This change only took about 1 minute, so it's 20 times cheaper! I hope you are getting the idea of how this works. we've already done two iterations of test-driven XP development in less time than it takes to read this. Now some excercises: 1. Find and install JUnit (hint, try http://xprogramming.com/) 2. Compile and run the above tests, and make sure they work. 3. Add a new unit test for "hello,world" making ["hello" "world"] 4. Update the code to make it pass the new test as well as the old ones. (hint, you should be getting closer to a general case with every change from now on. Remember to keep the code as simple as possible, refactor as necessary)
Frank, Excellent overview. If i may add. Done right, unit tests (in general, not XP) were given to programmers to test a unit of code, but would almost always be tossed. While the build and system tests usually migrated to regression testing for future systems, the effort that went into unit tests was wasted. One great reuse item of XP is the reuse of unit tests. == Additional benifit - it allows a programmer to have an acutual piece of working code when taking over an existing project ==