Streamable Fragments of Forward XPath

Abstract

Abstract. We present a query answering algorithm for a fragment of Forward XPath on Xml streams that we obtain by compilation to deterministic nested word automata. Our algorithm is earliest and in polynomial time. This proves the finite streamability of the fragment of Forward XPath with child steps, outermost-descendant steps, label tests, negation, and conjunction (aka filters), under the reasonable assumption that the number of conjunctions is bounded. We also prove that finite streamability fails without this assumption except if P=NP.

Early Nested Word Automata for XPath Query Answering on XML Streams

“…As a second contribution, we show how to compile XPath queries to eNwas by adapting the previous translation to dNwas from [10], while distinguishing selection and rejection states.…”

“…Unfortunately, this kind of earliest query answering is not feasible in polynomial time for XPath queries [6], as first shown by adapting counter examples from online verification [16]. A second argument is that deciding aliveness is more difficult than deciding XPath satisfiability [10], which is coNP-hard even for small fragments of XPath [4]. The situation is different for queries defined by deterministic nested word automata (Nwas) [1,2], for which earliest query answering is feasible with polynomial resources [20,11].…”

“…We hope that this unsatisfactory situation can be resolved in the near future by pushing existing automata techniques forwards [14,20,10,21].…”

“…The situation is different for queries defined by deterministic nested word automata (Nwas) [1,2], for which earliest query answering is feasible with polynomial resources [20,11]. Many practical XPath queries (without aggregation, joins, and negation) can be compiled into small Nwas [10], while relying on non-determinism for modeling descendant and following axes. This, however, does not lead to an efficient streaming algorithm.…”

“…Therefore, complex event processing for Xml streams has been studied for more than a decade [14,7,22,25,5,20,15,10,21]. Query answering for XPath is a basic algorithmic task on Xml streams, since XPath is a language hosted by the W3C standards Xslt and XQuery.…”