Transformation-based Indexing Techniques for Constraint Handling Rules

Multi-headed rules are essential for the expressiveness of Constraint Handling Rules (CHR), but incur considerable performance overhead. Current indexing techniques are often unable to address this problem—they require matchings to have particular form, or offer good run-time complexity rather than good absolute figures. We introduce two lightweight program transformations, based on term flattening, which improve the effectiveness of existing CHR indexing techniques, in terms of both complexity and constant factors. We also describe a set of complementary post-processing program transformations, which considerably reduce the flattening overhead. We compare our techniques with the current state of the art in CHR compilation, and measure their efficacy in K.U.Leuven CHR and CHRd.