Abstract

Object-Oriented design frameworks and design patterns have been introduced to document reusable and reliable designs. This paper addresses the problem of assessing the design quality of object oriented frameworks. Previous empirical studies show a correlation between certain types of coupling, cohesion, and complexity measures and error-prone software. A tightly coupled framework is highly interdependent which results in a poor design quality. Unified measures for coupling based on empirical studies of fault prone object-oriented code have recently been introduced and theoretically validated. In this paper, we propose a technique to determine coupling measures for high level designs of object-oriented frameworks expressed in UML. The coupling structure of a framework is explored in terms of what we called mutual class coupling and class coupling distributions where a class-coupling matrix is obtained for each coupling measure. We also show how the framework coupling structure affect the error-proneness, reusability and hence reliability of a design. Two framework designs are used as a case study; the first is produced by deploying design patterns and the second is obtained by reducing the classes of the first framework. Numerical coupling measures are presented and logically explained in the framework context.

Keywords: Coupling metric, Design Quality, Object-Oriented Design Frameworks, Design Patterns