View objects for relational databases

Abstract

Many application domains require database techniques for modeling and managing entities of arbitrary structure--the so-called complex objects. At the same time, a major incentive to exploit database management systems is the ability to support sharing of data among applications. In practice, however, these two objectives tend to conflict. Storing information in the form of complex objects can seriously inhibit sharing, since persistent objects bind application-dependent knowledge to the data. On the other hand, the relational model provides information sharing through, for example, the definition of views, but it lacks the expressive power to represent complex entities. This dissertation presents the view-object model as a first step toward reconciling the opposing objectives of object-oriented access to shared information. By combining the relational-database concept of view and the programming-language concept of object, the view-object model supports simultaneously abstract complex units of information and sharing of those units. Base information remains stored in a fully normalized relational database; this regular representation facilitates sharing. View objects are defined by complex, arbitrary structures that map cleanly to the underlying database through the use of a semantic data model--the structural model. Multiple view objects, specified over various parts of the database, hence offer different views of the data. Dynamic instantiation and update operations on view objects further support object-oriented manipulation of the information. In effect, the view-object model provides a unifying framework for merging database and knowledge-based techniques into an architecture for expert database systems. I have implemented this methodology in a prototype system called PENGUIN, which defines an object layer on top of a relational database system. For expert database systems that seek to exploit existing database technology, PENGUIN demonstrates that view objects serve as effective mediators, performing the knowledge-based task of transforming data into information.