Converting One Type-Based Abstract Domain to Another

The specific problem that motivates this paper is how to obtain abstract descriptions of the meanings of imported predicates (such as built-ins) that can be used when analysing a module of a logic program with respect to some abstract domain. We assume that abstract descriptions of the imported predicates are available in terms of some “standard” assertions. The first task is to define an abstract domain corresponding to the assertions for a given module and express the descriptions as objects in that domain. Following that they are automatically transformed into the analysis domain of interest. We develop a method which has been applied in order to generate call and success patterns from the CiaoPP assertions for built-ins, for any given regular type-based domain. In the paper we present the method as an instance of the more general problem of mapping elements of one abstract domain to another, with as little loss in precision as possible.