Evaluation of regular nonlinear recursions by deductive database techniques

摘要

Nonlinear recursion is one of the most challenging classes of logic programs for efficient evaluation in logic programming systems. We identify one popular class of nonlinear recursion, regular nonlinear recursion, and investigate its efficient implementation by a deductive database approach. The approach performs a detailed query binding analysis based on query information, constraint information and the structure of a recursion, selects an appropriate predicate evaluation order and generates an efficient query evaluation plan. Interesting query evaluation techniques, such as chain-following, chain-split, and constraint pushing, are developed for the efficient evaluation of different kinds of queries. Furthermore, the technique can be extended to the evaluation of regular nonlinear recursions in HiLog and F-logic programs. The study not only presents a method for the evaluation of regular nonlinear recursions in a declarative way but also demonstrates the power of the deductive database approach in the analysis and evaluation of sophisticated logic programs.