Attributed data for CHR indexing

Abstract

The overhead of matching CHR rules is alleviated by constraint store indexing. Attributed variables provide an efficient means of indexing on logical variables. Existing indexing strategies for ground terms, based on hash tables, incur considerable performance overhead, especially when frequently computing hash values for large terms. In this paper we (1) propose attributed data, a new data representation for ground terms inspired by attributed variables, that avoids the overhead of hash-table indexing, (2) describe program analysis and transformation techniques that make attributed data more effective, and (3) provide experimental results that establish the usefulness of our approach. © 2009 Springer Berlin Heidelberg.