Abstract
This paper presents an abstract view of mutation analysis. Mutation was originally thought of as making changes to program source, but similar kinds of changes have been applied to other artifacts, including program specifications, XML, and input languages. This paper argues that mutation analysis is actually a way to modify any software artifact based on its syntactic description, and is in the same family of test generation methods that create inputs from syntactic descriptions. The essential characteristic of mutation is that a syntactic description such as a grammar is used to create tests. We call this abstract view grammar-based testing, and view it as an interface, which mutation analysis implements. This shift in view allows mutation to be defined in a general way, yielding three benefits. First, it provides a simpler way to understand mutation. Second, it makes it easier to develop future applications of mutation analysis, such as finite state machines and use case collaboration diagrams. The third benefit, which due to space limitations is not explored in this paper, is ensuring that existing techniques are complete according to the criteria defined here.