An Automated Framework for Cost Reduction of Mutation Testing Based on Program Similarity

This paper presents an implementation and assessment of a framework named SiMut. The framework was introduced in a previous paper with the objective of helping reducing the cost for testing a program based on groups of similar programs previously tested with mutation. The implementation presented in the paper handles Java programs and includes a set of variants that relate to three types of program abstraction (original source code, processed source code, and internal complexity metrics), three similarity calculation strategies (clustering, information diversity, and plagiarism), and one mutation cost reduction approach (inspired by the One-Op mutation technique). Our evaluation encompasses 20 variant combinations, also referred to as SiMut configurations, and 35 small Java programs. A cross-comparison involving the formed clusters and a comparison with randomly formed clusters points to configurations that tend to reach high effectiveness in foreseeing the best mutation operators for untested programs.