From searching text to querying XML streams

Abstract

XML data is queried with XPath expressions, which are a limited form of regular expressions. New XML stream processing applications, such as content-based routing or selective dissemination of information, require thousands or millions of XPath expressions to be evaluated simultaneously on the incoming XML stream at a high, sustained rate. Conceptually, the XPath evaluation problem is analogous to the text search problem, in which one or several regular expressions need to be matched to a given text, but the number of regular expressions here is much larger, while the “text” is much shorter, since it corresponds to the depth of the XML stream. In this paper we examine techniques that have been proposed for XML stream processing, which are variations of either a non-deterministic or a deterministic finite automata (NFA and DFA).F or the latter, we describe a series or theoretical results establishing lower and upper bounds on the number of DFA states for sets of XPath expressions.