Evolutionary generation of unique input/output sequences for class behavioral testing

Object-oriented software is composed of classes. Their behaviors are usually modeled with state diagrams or finite state machines (FSMs). Testing classes is regarded as testing FSMs in which unique input/output (UIO) sequences are widely applied. The generation of UIO sequences is shown to be an undecidable problem. For these problems, genetic algorithms (GAs) may offer much promise. This paper reports some primary results of on-going research on evolutionary testing classes. First, we explain how to define UIO sequence generation as a search problem, and then describe adapting genetic algorithms to generating UIO sequences. Special issues of using genetic algorithms such as solution representation, validity checking and fitness definition are discussed in detail. Primary experiments confirm the applicability and feasibility of applying GAs to UIO sequence generation.