Restructuring for Large Data Bases: Three Levels of Abstraction

The development of a powerful restructuring function involves two important components - the unambiguous specification of the restructuring operations and the realization of these operations in a software system. We direct our efforts to the first component in the belief that a precise specification will provide a firm foundation for restructuring algorithms and implementations. This paper defines completely the semantics of the restructuring of tree structured data bases. The delineation of the restructuring function is accomplished by formulating three different levels of abstraction, with each level of abstraction representing successively more detailed semantics of the function. At the first level of abstraction, the schema modification, three types are identified: naming, combining and relating, and are further divided into eight schema operations. The second level of abstraction, the instance operations, constitutes the transformations on the data instances. They are divided into group operations such as replication, factoring, union and group relation operations such as collapsing, refinement, fusion etc. The final level, the item values operations, includes the actual item operations such as copy value, delete value, or create a null value.