A Bottom-up Strategy for Query Decomposition

Abstract

In order to access data from various different data repositories, in global-as-view approaches an input query is decomposed into several subqueries. Normally, this decomposition is based on a set of mappings, which describe the correspondence of data elements between a global schema and local ones. However, building mappings is a difficult task, especially when the number of participating local schemas is large. In our approach, an input query is automatically decomposed into subqueries without using mappings. An algorithm is proposed to transform a global path expression (e.g., an XPath query) into local path expressions (e.g., XPath queries) executable in local schemas. This algorithm transforms parts of a path expression from right to left. This transformation is applied from the bottom to the top of a tree and depends on structures of local schemas. Compared to top-down approaches as by Lausen and Marron (IM), our bottom-up approach can be more efficient. Even in the worst case, the time complexity of our algorithm can be n times better than that of IM, where n is the number of parts in a global query. In the best case, for a k-ary tree of height h, the time complexity of our algorithm is T(n,k,h)= min(n,h), whereas that of LM we have found is T(n,k,h,)=n*(kh+1-1)/(k-1) This can reduce to a large extent the time for forming subqueries for local (e.g., XMI) schemas.