Approximate Pruning in Tabled Logic Programming

Abstract

Abstract. Pruning provides an important tool for control of nondeterminism in Prolog systems. Current Tabled Prolog systems improve Prolog's evaluation strategy in several ways, but lack satisfactory support for pruning operations. In this paper we present an extension to the evaluation mechanism of Tabled Prolog to support pruning. This extension builds on the concept of demand to select tables to prune. In particular, we concentrate on systems based on SLG resolution. A once operator is described, which approximates demand-based pruning, providing for an efficient implementation in the XSB system.

Just enough tabling

“…If pruning takes place as proposed by Castro and Warren [2] the action is similar, because none of the tables is demanded at this point. In either case, if a call to one of these tables is encountered later in the computation, all execution performed so far for these tables will need to be repeated.…”

“…The approach of Castro and Warren to approximate pruning of tabled logic programs [2] is closer in spirit to that of Prolog where a pruning operator forgets computations and their alternatives which lie in its scope. Hence, it is not surprising that the pruning mechanism of [2] does not cater for re-activation of tables after pruning.…”

“…Typically, after completion, space for generators and consumers is reclaimed but information about calls and answers to tabled predicates is persistent and only deallocated through explicit programmer intervention, regardless of whether there is anticipated demand for these tables. In a tabled system that supports some form of pruning, either via Prolog's cut ( [15,7]) or as suggested by Castro and Warren [2], tables that are still incomplete and lie in the scope of a pruning operator are typically abolished when pruning occurs. This is because tabling engines do not maintain enough information to allow them to continue the process of tabled resolution from an arbitrary execution point.…”

Supporting Pruning in Tabled LP

“…To the best of our knowledge, there are five previous attempts to incorporate pruning operations in tabled LP [11,12,9,13,10]. [11] works under the dynamic reordering of alternatives (DRA) technique [14].…”

“…Demanded tables [13] implements a version of once/1. In this work, calls which are being consumed by external consumers (demanded table) are not pruned, which makes it necessary to perform runtime analysis to detect if a generator call is being demanded.…”