Trie Indexes for Efficient XML Query Evaluation

As the number of applications that rely on XML data increases, so does the need for performing efficient XML query evaluation. A critical part of the solution involves providing new techniques for designing XML indexes and lookup algorithms. In this paper, we leverage the results of our research on coupling the partitions induced by fragments of XPath algebra and those induced by the structural properties of an XML document to lead the design of the N [k] and P [k]-Trie indexes for XML. We present the rationale behind our approach, and detail the structure of the indexes, their features, and algorithms for evaluating XPath queries with index-only plans. We show that the P [k]-Trie indexes are capable of answering XPath queries with arbitrarily complicated predicates using index-only plans and improving query performance in orders of magnitude over the A[k]-index.