Write Tests for People
You are writing automated tests for some or all of your production code. Congratulations! You are writing your tests before you write the code? Even better!! Just doing this makes you one of the early adopters on the leading edge of software engineering practice. But are you writing good tests? How can you tell? One way is to ask "For whom am I writing the tests?" If the anwser is "For me, to save me the effort of fixing bugs." or "For the compiler so they can be executed!" then the odds are you aren't writing the best possible tests. So who should you be writing the tests for? For the person who's trying to understand your code.
Good tests act as documentation for the code they are testing. They describe how the code works. For each usage scenario the test(s):
- describe the preconditions that must be satisfied,
- illustrate how the software is invoked
- describe the expected results.
Different usage scenarios will have slightly different versions of each of these. The person trying to understand your code should be able to look at a few tests and by comparing these three parts of the tests in question, be able to see what causes the software to behave differently. Each test should clearly illustrate the cause and effect relationship between these three parts. This implies that what isn't visible in the test is just as important as what is visible. Too much code in the test distracts the reader with unimportant trivia. Whenever possible hide such trivia behind meaningful method calls (the Extract Method refactoring is your best friend.) And make sure you give each test a meaningful name that describes the particular usage scenario.
It is also a good idea to test your tests. You can verify they detect the errors you think they detect by inserting those errors into the production code (your own private copy that you'll throw away, of course). Make sure they report errors in a helpful and meaningful way. You should also verify that your tests speak clearly to a person trying to understand your code. The only way to do this is to have someone who isn't familiar with your code read your tests and tell you what they learned about the. Listen carefully to what they say; if they didn't understand something clearly it probably isn't because they aren't very bright. It is more likely because you weren't very clear. (Go ahead and reverse the roles by reading their tests!)