Efficient algorithms for evaluating xpath over streams

Abstract

In this paper we address the problem of evaluating XPath queries over streaming XML data. We consider a practical XPath fragment called Univariate XPath, which includes the commonly used '/' and '//' axes and allows * -node tests and arbitrarily nested predicates. It is well known that this XPath fragment can be efficiently evaluated in O(|D||Q|) time in the non-streaming environment [18], where |D| is the document size and |Q| is the query size. However, this is not necessarily true in the streaming environment, since streaming algorithms have to satisfy stricter requirement than non-streaming algorithms, in that all data must be read sequentially in one pass. Therefore, it is not surprising that state-of-the-art stream-querying algorithms have higher time complexity than O(|D||Q|).In this paper we revisit the XPath stream-querying problem, and show that Univariate XPath can be efficiently evaluated in O(|D||Q|) time in the streaming environment. Specifically, we propose two O(|D||Q|)-time stream-querying algorithms, LQ and EQ, which are based on the lazy strategy and on the eager strategy, respectively. To the best of our knowledge, LQ and EQ are the first XPath stream-querying algorithms that achieve O(|D||Q|) time performance. Further, our algorithms achieve O(|D||Q|) time performance without trading off space performance. Instead, they have better buffering-space performance than state-of-the-art stream-querying algorithms. In particular, EQ achieves optimal buffering-space performance. Our experimental results show that our algorithms have not only good theoretical complexity but also considerable practical performance advantages over existing algorithms.

Efficient Streaming Algorithms for Tree Matching Problems

“…In the experiments, we have tested four methods on the tree matching: -TwigM [12] , -Gou's method [23] , -the method for the unordered tree matching (discussed in this paper, chen-1 for short), and -the method for the ordered tree matching (discussed in this paper, chen-2 for short). The theoretical computational complexities of these methods are summarized in Table 1.…”

“…If in Q the nodes are labeled with different tag names, BS is in the order of |T|. (See [23]). However, if different nodes in Q are labeled with the same tag name, BS might be bounded by O(|T|⋅|Q|).…”

“…One is for the so-called XML streaming environment and the other is index-based. A great many strategies have been proposed to evaluate XPath queries in an XML streaming environment [2,12,18,23,27,32,33,39,40] . The methods discussed in [2,27] are based on finite state automata (FSA), but only able to handle single path queries, i.e., a query containing branching cannot be processed, as observed in [39].…”

“…The methods discussed in [32,33] [12]. The algorithm discussed in [23] greatly improves TwigM, running in O(|Q|⋅|T|) time and O(|Q|⋅r) caching space.…”

Streamable Fragments of Forward XPath

“…Query answering algorithms for XPath on Xml streams received much interest in the database and document processing communities [2,24,3,22,5,12,29,6,21] and are currently in the focus of the W3C working groups on Xslt and Xproc [14]. A little surprisingly, the topic is far from being settled given the large remaining gap between known streamable and non-streamable fragments.…”