Test Set Diameter: Quantifying the Diversity of Sets of Test Cases
This paper introduces the concept of test suite latency. The more latent a test suite, the more it is possible to repeatedly select subsets that achieve a test goal (such as coverage) without re-applying test cases. Where a test case is re-applied it cannot reveal new information. The more a test suite is forced to re-apply already applied test cases in order to achieve the test goal, the more it has become `worn out'. Test suite latency is the flipside of wear out; the more latent a test suite, the less prone it is to wear out. The paper introduces a theory of test suite latency. It presents results from the empirical study of latency, highlighting the need for latency enhancement. The paper also introduces a strategy and algorithms for improving latency and an empirical study of their effectiveness. The results show that local search is effective at improving the latency of a test suite.