Some Modifications of Bucket-Based Algorithms for Query Rewriting Using Views

Query rewriting using views is an important topic in data integration. A number of rewriting algorithms, such as the SVB algorithm, the MiniCon algorithm and the inverse rules algorithm, have been developed. All the algorithms can generate a maximally-contained rewriting of a given query, which is the union of a set of conjunctive rewritings contained in the given query. In this paper, we first argue that the condition for forming a shared-variable bucket in the SVB algorithm can be modified in the case where a shared variable in a query is mapped to a distinguished variable that is also a join attribute in a view. Under the modified condition, we may create more shared-variable buckets so that fewer rewritings can be generated than the SVB algorithm. Second, the SVB algorithm does not handle a constant in a view properly in the case where a shared variable of a query is mapped to a constant of a view. We propose to use a pseudo shared-variable bucket to address this issue. The only difference between the SVB algorithm and the MiniCon algorithm is that the latter considers a head homomorphism on a view. However, the head homomorphism on a view is not related to the condition we intend to modify in this paper. The modifications we present are also applicable to the MiniCon algorithm.