Bridging the Gap Between Design and Implementation of Component Libraries

Object{oriented design is usually driven by three main reu-sability principles: step{by{step design, design for reuse and design with reuse. However, these principles are just partially applied to the subsequent object{oriented implementation, often due to eeciency constraints , yielding to a gap between design and implementation. In this paper we provide a solution for bridging this gap for a concrete framework, the one of designing and implementing container{like component libraries, such as STL, Booch Components, etc. Our approach is based on a new design pattern together with its corresponding implementation. The proposal enhances the same principles that drive the design process: step{ by{step implementation (adding just what is needed in every step), implementation with reuse (component implementations are reused while library implementation progresses and component hierarchies grow) and implementation for reuse (intermediate component implementations can be reused in many diierent points of the hierarchy). We use our approach in two diierent manners: for building a brand{new container{like component library, and for reengineering an existing one, Booch Components in Ada95. 1 Motivation Since McIlroy proposed in 1969 the notion of software catalogue McI69], compo-nent{based software development (CBSD) has become without any doubt one of the most important software development paradigms. The key point behind this paradigm is the process of reusing components from standard software catalogues. Component reuse provides many advantages, remarkably software production hastening, software quality improvement and software maintenance cost decrease. One of the most valuable contributions in CBSD is object{oriented (OO) technology. Basic concepts such as inheritance, polymorphismand dynamic binding , and others built on top of them, such as design patterns GHJ+96], had a strong impact on this paradigm. In fact, as Meyer remarks Mey97], OO technology makes possible, for the rst time, the idea of turning academic McIlroy's vision of software development into a component{based industry become real. 2 Unfortunately, despite of this ongoing success, we have not still reached the ultimate McIlroy's goal: making software components equivalent to other engineering component types, as chips, bricks and valves. One of the reasons behind this fact has to be with the diiculty of turning designs into implementations. The OO design process is usually driven by three main reusability principles, namely step{by{step design, design for reuse and design with reuse. However, these principles are often just partially applied to the subsequent OO implementation. Therefore, a gap between design and implementation appears with respect to these reusability principles. This fact …