General variable strength t-way strategy supporting flexible interactions

To ensure conformance and establish quality, software testing is an integral part in software engineering lifecycle. However, because of resource and time-to-market constraints, testing all exhaustive possibilities is impossible in nearly all practical testing problems. Considering the aforementioned constraints, much research now focuses on a sampling technique based on interaction testing (termed as t-way strategy). Although helpful, most t- way strategies (e.g. AETG, In-Parameter-Order General (IPOG), and GTWay) assume that all parameters have uniform interaction. In reality, the interaction among parameters is rarely uniform. Some parameters may not even interact, wasting the testing efforts. As a result, a number of newly developed t-way strategies that consider variable-strength interaction based on input-output relationships have been developed, e.g. Union, ParaOrder, and Density. Although useful, these strategies often suffer from lack of optimality in terms of the generated test size. Furthermore, no single strategy is dominant because the optimal generation of t-way interaction test suite is considered an Nondeterministic Polynomial (NP) hard problem. Motivated by the above-mentioned challenges, this paper proposes and implements a new strategy, called General Variable Strength (GVS). GVS has been demonstrated, in some cases, to produce better results than other competing strategies.