Automated acquisition and refinement of reusable software design components

Abstract

The importance of reusing software components in new software developments has long been recognized. A reason for the current low level of software reuse stems from the difficulties of obtaining appropriate components for populating a reuse library.

This thesis presents a framework for a hybrid software design environment that supports the reuse of design schemas and cases. At the heart of this framework is a knowledge-based approach for automating the acquisition and refinement of software components for its reuse library. A prototype system, CAReT, has been implemented to demonstrate the viability of the approach.

CAReT is based on a bottom-up, case-based approach to domain analysis, where design cases for existing systems within an application domain are examined to determine common pattern of operations and objects. The correspondences between features are established using domain knowledge, where available. Otherwise the correspondence identification is decided by using a similarity metric to compute the degree of closeness between features. The common features are then abstracted. This same mechanism for determining correspondence underlies CAReT's analogy-based retriever, which retrieves software components that satisfy a reuser's query.

An application domain life cycle model is introduced to explain the implications of a domain's maturity on efforts to acquire reusable software components. In particular, the impact on design schema derivation is examined.