Logic programming: ideal vs practice

Abstract

The logician who encounters logic programming expects it to involve 'using predicate logic as a programming language'. He is soon disillusioned. He discovers that it really means 'programming in PROLOG', which in tum seems to mean 'programming in a procedural language combining a peculiar syntax with an unusual flow of control that runs counter to certain basic assumptions of logic'. The baffled logician is left wondering how PROLOG could have come to be viewed as a logic programming language. In this paper we compare Gabbay's logicprogramming ideal with logic programming in practice and show how they may be reconciled. The reconciliation has implications for the teaching of logic programming and throws some light on the possible nature of future implementations. A crucial role is played in the discussion by the notion of feasibility. It is argued that, as yet, no appropriate theoretical framework for analysing the efficiency of proof procedures exists, but that some indication of what such a framework should look like can be gleaned from practice.