A Framework for Ad-Hoc Type Inference

Other widely used languages have been created less by design and more by coincidence and compromise. It seems therefore that the question of type inference for such languages could be infeasible or should at least permit only ad-hoc solutions.

In this paper we argue the existence of a uniform framework for developing type inference algorithms, even for seemingly ad-hoc languages. This framework provides useful generalizations of well-known concepts related to type inference. It is, however, by no means a universal algorithm. In fact, the essential algorithmic problem must generally be solved from scratch.