It takes about two seconds to determine whether or not someone used TDD to derive a set of unit tests. While it’s possible for someone to write as high-quality unit tests using TAD (test-after development), it just doesn’t happen. Invariably I see many of the same problems, most significant of which is that the TAD tests don’t cover nearly enough of the cases.
Most of the time, developers doing TAD simply refuse to follow the single-responsibility principle (SRP) at the class or method level. (Despite bizarre claims by Jeff Atwood and Joel Spolsky, asking people to push in the direction of Uncle Bob’s five principles of class design is a reasonable and sensible thing to do.) They build large classes and large methods–not small, cohesive, potentially reusable units–and then complain about the difficulty of testing them.
My simple answer for now is, “How can you expect to write unit tests when you don’t have any units?”
