From Monomorphic to Polymorphic Well-Typings and Beyond

Abstract

Type information has many applications; it can e.g. be used in optimized compilation, termination analysis and error detection. However, logic programs are typically untyped. A well-typed program has the property that it behaves identically on well-typed goals with or without type checking. Hence the automatic inference of a well-typing is worthwhile. Existing inferences are either cheap and inaccurate, or accurate and expensive. By giving up the requirement that all calls to a predicate have types that are instances of a unique polymorphic type but instead allowing multiple polymorphic typings for the same predicate, we obtain a novel strongly-connected-component-based analysis that provides a good compromise between accuracy and computational cost.

Towards Parameterized Regular Type Inference Using Set Constraints

“…Bruynooghe, Gallagher, Humbeeck, and Schrijvers [3,21] develop analyses for type inference which are also based on the set constraints approach. However, their analyses infer well-typings, so that the result is not an approximation of the program success set, as in our case.…”

“…In consequence, their algorithm is simpler, mainly because they do not need to deal with intersection. In [21] the monomorphic analysis of [3] is extended to the polymorphic case, using types with an expressiveness comparable to our parameterized types.…”

“…1. Parametric vs parameterized regular types which are conservative approximations of the meaning of predicates, and hence, overapproximations of the success set (in contrast to the approach of inferring well-typings, as in, e.g., [3,21], which may differ from the actual success set of the program). Type inference via set constraint solving was already proposed in [12,15].…”

A Transformational Approach for Proving Properties of the CHR Constraint Store

“…In Logic Programming (LP), many tools are available for performing such analyses [2,5,12]. Hence, instead of building analyses for CHR from scratch, it is interesting to explore whether one can define transformations from CHR to Prolog and reuse existing analysis tools for Prolog to obtain properties about the constraints that are in the CHR constraint store during computations.…”