Data abstraction mechanisms in object-oriented programming

• Main Author: Li, Haitao
• Other Authors: Mili, Hafedh (advisor)
• Format: Others
• Language: en
• Published: McGill University 1993
• Subjects: Computer Science.
• Online Access: http://digitool.Library.McGill.CA:80/R/?func=dbin-jump-full&object_id=69723

Existing OO modeling methodologies prescribe notations, processes, and guidelines that, if followed, ensure that analysis-level OO model reflect application semantics. As we move into design, implementation-level considerations may distort analysis-level models, and the transition is seamless no more. In this thesis, we describe data definition facilities in SoftClass--an experimental CASE tool for software reuse--that aim at maintaining the integrity of application data models throughout the development lifecycle, while maximizing opportunities for code reuse. In SoftClass, analysis-level data models describe application-semantics and are organized in an inheritance hierarchy based on shared application-semantics. At the design-level, we maintain two kinds of data models: (1) generic data structures, used as implementation templates for analysis-level models, and organized along "implementation inheritance" hierarchies, and (2) realizations of analysis-level application models, which consist of mapping an analysis-level model to a generic data structure. Design-level representations of application objects may be seen as belonging to two independent hierarchies, and we show how each hierarchy offers some opportunities for reuse. We shoe how data abstraction supports a high-level program design language that is both easy-to-use and that supports some design validation. We conclude by outlining directions for further research.