Impact of mutation operators on mutant equivalence

The presence of equivalent mutants is a recurrent source of aggravation in mutation-based studies of software testing, as it distorts our analysis and precludes assertive claims. But the determination of whether a mutant is equivalent to a base program is undecidable, and practical approaches are tedious, error-prone, and tend to produce insufficient or unnecessary conditions of equivalence. We argue that an attractive alternative to painstakingly identifying equivalent mutants is to estimate their number. This is an attractive alternative for two reasons: First, in most practical applications, it is not necessary to identify equivalent mutants individually; rather it suffices to know their number. Second, even when we need to identify equivalent mutants, knowing their number enables us to single them out with little to moderate effort.