Finite segmentation for XML caching

Abstract

XML data processing often relies on basic relations between two XML fragments like containment, subset, difference and intersection. Fast calculation of such relations based only on the representing XPath expression is known to be a major challenge. Recently XML patterns have been introduced to model and identify handy subclasses of XPath. We present the concept of ST-pattem segments that uses sets of adapted tree patterns in order to describe a finite and complete partitioning of the XML document's data space. Based on such segmentations, we present a fast evaluation of XML relations and show how to compute a set of patterns for an optimal segmentation based on frequent XPath queries. © 2005 by International Federation for Information Processing.