Conceptual Foundations for the Design Transformation of Procedural Software to Object-Oriented Architecture

The movement from algorithmic decomposition, structured design and procedural implementation towards objectoriented design and programming has resulted in a growing need to restructure/re-engineer old programs which have a poor architecture in order to reap the benefits of this new technology. It is more economic to restructure existing software into an object-oriented topology by identifying objects within the subject system rather than to redesign the software starting from the requirements. Such object identification relies upon program analysis and a study of inter-component relationships and dependencies. This paper initially provides a conceptual foundation for migration from a procedural to an object-oriented software architecture for legacy software, which relies upon a view of the software life cycle, in which all software development is considered to be an evolutionary activity with re-engineering/restructuring as an important process applied repeatedly to the artefacts of development at various stages in this evolution. We then introduce a novel approach of viewing a procedural program from an object-oriented perspective in which there is single “god” class, which has a large number of global variables and methods, supported by several user defined classes which have no behaviour, only data definitions. This view, coupled with repeated restructuring, enables a seamless migration of behaviour from the god class to other classes using design transformation methods. There is a distinction between the processes of translation and transformation, processes that we contend are orthogonal to each other, which thereby enables us to concentrate on behaviour-preserving design transformations in order to migrate from one design to another, leaving the language issues to be handled by translators. Finally, the paper describes some basic design transformations which preserve system behaviour over the partitioning of classes. A CASE tool ‘RESTRUCT’ which is based on these ideas is being implemented to test the ideas presented in this paper.