A Novel Decomposition of Multilevel Relations into Single-Level Relations

In this paper we give a new decomposition algorithm that breaks a multilevel relation into single-level relations and a new recovery algorithmwhich reconstructs the original multilevel relation from the decomposed single-level relations. There are several novel aspects to our decomposition and recovery algorithms which provide substantial advantages over previous proposals: (1) Our algorithms are formulated in the context of an operational semantics for multilevel relations, dened here by generalizing the usual update operations of SQL to multilevel relations. (2) Our algorithms, with minor modi cations, can easily accommodate alternative update semantics which have been proposed in the literature. (3) Our algorithms are e cient because recovery is based solely on union-like operations without any use of joins. (4) Our decomposition is intuitively and theoretically simple giving us a sound basis for correctness. In this paper we also argue that some of the alternate update semantics which have been proposed for multilevel relations should be available as options, but should certainly not be made an integral part of the data model.