Counting-based normalization for multiple linear recursions

Abstract

Normalization is a powerful query independent compilation and optimization technique for complex linear recursions in deductive databases [6, 7, 10, 8, 4]. It transforms a linear recursion into an n-chain recursion which consists of n chain predicates among which there is no shared variables. Normalization facilitates capturing more bindings and doing quantitative analysis, hence generating efficient query processing plans. However, the existing normalization methods are applicable only to single linear recursions. In this paper, we revise the concept of chain and n-chain recursion, and propose a novel method, called counting-based transformation, to transform a general multiple linear recursive program to single n-chain recursion. Based on this method, the existing evaluation methods which are applicable to single linear recursions can be applied also to multiple linear recursions.