A Collection Model for Data Management in Object-Oriented Systems

This thesis addresses the question of how to provide data management services in object-oriented systems with reliable persistent object stores. It proposes an object data model, called the collection model, which serves as a foundation for the construction of such services. The collection model is general in that it is independent of any particular implementation platform. In part, this independence is achieved through the separation of the data model from the underlying type model. There are two components of the collection model - a structural model, BROOM, and an operational model based on an algebra of collections. The structural model is semantically rich and exhibits properties of both the entity-relationship and semantic data models. Hoary collections are used to represent entity categories and binary collections to represent relationships between entities. Classification structures are based on the notion of a. collection family which represents various forms of conceptual dependencies among the collections of a. family. The requirements for supporting the various forms of evolution in object-oriented database systems are presented. An extension to the collection model is proposed to support object evolution whereby objects can migrate within classification structures. Two existing realisations of the collection model are described. One is a. prototype, single-user system implemented in Prolog. The other forms the basis of the Object Data. Management Services of the Comandos platform for distributed, object-oriented applications. A general approach to object data model design, specification and realisation is advocated. In particular, a metacircular description of the collection model is used as an intermediate form of data model specification. This metacircular description is then transformed into a formal specification in the Z language