Database abstractions : an interpretation and implementation

In the relational data model, the database consists of a potentially large number of relations. No means of logically grouping the relations exists. In [3,4,5,6], Smith and Smith introduce two abstractions, aggregation and generalization, as a way to build hierarchical relationships among relations at the conceptual level. These abstractions give the database user better understanding and control over the conceptual design of the whole database. At the same time, they preserve the properties of the relational model:

A relation is still a table and consequently the relation operations still can be performed on these relations; normalization theory still applies to tables so that tables can be transformed to various normal forms which are free of undesirable redundancy and update anomalies.

This thesis is based on the work by Smith and Smith [3,4,5,6] on database abstractions. It discusses problems with their approach, justifies changes made to their syntax and semantics to overcome those problems, and allows for a broader class of interpretations of what they term "relational Several invariant" properties of hierarchies of relations. possible implementations of aggregation and generalization abstractions are considered before the actual implementation based on the relational database system SUR[7] is discussed. The implementation is not proved correct formally but is demonstrated using examples in a SITBOL program designed to run on a DEC-10 computer.