Function Materialization in Object-Oriented Databases

Abstract

This paper considers aspects of object-oriented database support for engineering design. In particular we consider function materialization, that is the use of stored, precalculatedfunction results to optimize the return of a function result. We review die materialization of functions in the relational data model and an object-oriented data model implementation. We present a new approach to function materialization in object-oriented databases, as implemented in our class library TOMS. This approach is fundamentally different to a previous approach, due to the context in which we see function materialization being used; an environment where complex engineering design objects evolve through a series of versions. The objects storing the precalculated function results - MFOs, will themselves exist in several versions. We describe the means of managing these complex version evolution configurations, and detail the expression of relationships which exist between MFOs and dependent objects. We go on to describe the way in which rematerialization of functions after changes to objects within a database, can be supported by object versioning in a time efficient manner. The designer must carefully consider the trade-off between improved function result response times and the storage costs of maintaining complex version hierarchies. Finally we show that MFOs may be extracted from their associated version evolution trees and checked in to a public database in a hierarchically partitioned engineering design database environment.