Ecological Partial Deduction : Preserving Characteristic

A partial deduction strategy for logic programs usually uses an abstraction operation to guarantee the niteness of the set of atoms for which partial deductions are produced. Finding an abstraction operation which guarantees niteness and does not loose relevant information is a diicult problem. In earlier work Gallagher and Bruynooghe proposed to base the abstraction operation on characteristic paths and trees. A characteristic tree captures the relevant structure of the generated partial SLDNF-tree for a given goal. Unfortunately the abstraction operations proposed in the earlier work do not always produce more general atoms and do not always preserve the characteristic trees. This problem has been solved for purely determinate unfolding rules and deenite programs in 12, 13] by using constraints inside the partial deduction process. In this paper we propose an alternate solution which achieves the preservation of characteristic trees for any unfolding rule, normal logic programs (it can even handle some built-in's if so desired) and without adding constraints to the partial deduction process (making the re-use of existing unfolding techniques very simple). We thus provide a powerful, generally applicable and elegant abstraction operation for partial deduction providing a ne-grained and terminating control of polyvariance.