Automatic inference of abstract type behavior

Abstract

Type hierarchies are an integral part of the object oriented software reuse machinery. Software flexibility can be increased through type inheritance which, if used in accordance with Liskov Substitution Principle (LSP) enables safe object substitution.Assuming that formal specifications are available for a set of subtypes, we present our early doctoral research on the automatic inference of an extended deterministic finite automaton that describes the legal usage of abstract supertypes and ensures the behavioral subtyping relation as defined by the Liskov Substitution Principle (LSP). We obtain the supertype interface automata by incrementally exploring the specification of the subtypes, unifying correlated subtype fields, simplifying predicates through quantification, and finally creating new model fields that we associate with the remaining predicates.The inferred automaton is simulated by the behavior of each subtype and can be used for safe hierarchy extension, verification of new hierarchy clients, and emphasis of LSP non-compliant methods.