Dynamic Analysis of Algebraic Structure to Optimize Test Generation and Test Case Selection

Where no independent specification is available, object-oriented unit testing is limited to exercising all interleaved method paths, seeking unexpected failures. A recent trend in unit testing, that interleaves dynamic analysis between each test cycle, has brought useful reductions in test-set sizes by pruning redundant prefix paths. This paper describes a dynamic approach to analyzing the algebraic structure of test objects, such that prefix paths ending in observer or transformer operations yielding unchanged, or derived states may be detected and pruned on-the-fly during testing. The fewer retained test cases are so close to the ideal algebraic specification cases that a tester can afford to confirm or reject these cases interactively, which are then used as a test oracle to predict many further test outcomes during automated testing. The algebra-inspired algorithms are incorporated in the latest version of the JWalk lazy systematic unit testing tool suite, which discovers key test cases, while pruning many thousands of redundant test cases.