Bag Equivalence of XPath Queries

摘要

When a query is evaluated under bag semantics, each answer is returned as many times as it has derivations. Bag semantics has long been recognized as important, especially when aggregation functions will be applied to query results. This paper is the first to focus on bag semantics for XPath queries. In particular, the problem of bag-equivalence of a large class of XPath queries (modeled as tree patterns) is explored. The queries can contain unions, branching, label wildcards, the vertical child and descendant axes, the horizontal following, following-sibling and immediately-following sibling axes, as well as positional (i.e., first and last) axes. Equivalence characterizations are provided, and their complexity is analyzed. As the descendent axis involves a recursive relationship, this paper is also the first to address bag equivalence over recursive queries, in any setting.