The information preserving properties of relational database transformations

Abstract

Several areas of relational database research rely on techniques for transforming the schema (intention) and underlying occurrence structures (extension) of a database. In each of these areas, it is imperative to know what from the original database is preserved by the restructuring operation. This paper examines the extent to which an important class of relational database transformations preserves information. This is accomplished in two steps. First, a formal characterization of information content is presented based on the set of functional and nonfunctional associations among attributes of a relational database. Second, formal conditions are presented indicating the extent to which these associations are preserved. Based on these conditions, the preserving properties of several well-known relational transformations are described. Some problems in maintaining these properties under updates are also discussed.