XML Reconstruction View Selection in XML Databases: Complexity Analysis and Approximation Scheme

Query evaluation in an XML database requires reconstructing XML subtrees rooted at nodes found by an XML query. Since XML subtree reconstruction can be expensive, one approach to improve query response time is to use reconstruction views - materialized XML subtrees of an XML document, whose nodes are frequently accessed by XML queries. For this approach to be efficient, the principal requirement is a framework for view selection. In this work, we are the first to formalize and study the problem of XML reconstruction view selection. The input is a tree $T$, in which every node $i$ has a size $c_i$ and profit $p_i$, and the size limitation $C$. The target is to find a subset of subtrees rooted at nodes $i_1,\cdots, i_k$ respectively such that $c_{i_1}+\cdots +c_{i_k}\le C$, and $p_{i_1}+\cdots +p_{i_k}$ is maximal. Furthermore, there is no overlap between any two subtrees selected in the solution. We prove that this problem is NP-hard and present a fully polynomial-time approximation scheme (FPTAS) as a solution.