Fast and Tiny Structural Self-Indexes for XML

TL;DR

This paper introduces a compact, efficient index for grammar-compressed XML trees that accelerates structural XPath queries and serialization, outperforming previous methods in speed and space efficiency.

Contribution

It presents a fully-fledged index over grammar-compressed XML trees enabling faster query execution and serialization without decompression, improving performance significantly.

Findings

XPath count queries are often two orders of magnitude faster over the index.

Serialization of XML results is 2-3 times faster than previous systems.

The index enables efficient execution of arbitrary tree algorithms with minimal slowdown.

Abstract

XML document markup is highly repetitive and therefore well compressible using dictionary-based methods such as DAGs or grammars. In the context of selectivity estimation, grammar-compressed trees were used before as synopsis for structural XPath queries. Here a fully-fledged index over such grammars is presented. The index allows to execute arbitrary tree algorithms with a slow-down that is comparable to the space improvement. More interestingly, certain algorithms execute much faster over the index (because no decompression occurs). E.g., for structural XPath count queries, evaluating over the index is faster than previous XPath implementations, often by two orders of magnitude. The index also allows to serialize XML results (including texts) faster than previous systems, by a factor of ca. 2-3. This is due to efficient copy handling of grammar repetitions, and because materialization is totally avoided. In order to compare with twig join implementations, we implemented a materializer which writes out pre-order numbers of result nodes, and show its competitiveness.