XML Prefiltering as a String Matching Problem

We propose a new technique for the efficient search and navigation in XML documents and streams. This technique takes string matching algorithms designed for efficient keyword search in flat strings into the second dimension, to navigate in tree structured data. We consider the important XML data management task of prefiltering XML documents (also called XML projection) as an application for our approach. Different from existing prefiltering schemes, we usually process only fractions of the input and get by with very economical consumption of both main memory and processing time. Our experiments reveal that, already on low-complexity problems such as XPath filtering, in-memory query engines can experience speed-ups by two orders of magnitude.