An interpreter with lazy evaluation for Prolog with functions

Abstract

If Horn clause logic is enriched with functions that are specified via (conditional) equations, the well-known SLD-resolution algorithm has to be extended by a semantic unification. An implementation of such a language has to be designed very carefully to reach the performance of modern functional languages like ML or Miranda. It is also desirable to have a lazy evaluation mechanism, at least for the functional portion of the language.

We define a syntax and informal semantics for logic programs with functions. The evaluation of function calls is lazy. The language is designed to be practically useful (e. g. all predefined predicates of the host Prolog can be used in the definition of functions and predicates) and to have a close co-operation of the logical and functional parts of a program.

A naive interpreter is described to show certain kinds of problems minimizing the number of reduction steps. Then our approach is presented which includes an interpreter and a compiler (both written in Prolog) to transform the clauses into a form suitable for efficient interpretation. A special term representation is essential for an efficient execution of the reduction steps. This term representation may be characterized as delayed unification.

Finally we argue that our term representation is very suitable for implementation within the Prolog interpreter itself, i. e. not on the meta-level. This will lead to a really efficient implementation of the combined logic and functional programming paradigm.

A listing of the interpreter is included in the appendix.