A class library management system for object-oriented programming

Abstract

Object-oriented (00) software development has been touted to promote and enhance the reuse of existing software. This advantage, together with its emphasis on concepts like data abstraction, information hiding, encapsulation and inheritance, provides significant advantages as it allows software developers to tackle complex problems several orders of magnitudes higher than existing programming technologies can afford. However, if this ability to manage complexity is to become a reality, an 00 development environment must learn to manage its class libraries efficiently so that its developers can reuse the developed libraries and not have to re-invent what is already available. The problem becomes particularly acute when the number of classes increases to a certain level in an environment such that retrieval of information on existing classes becomes impossible. This phenomenon frequently occurs in developers of CASE tools since the level of problem they have to tackle is much more complex and consequently their repositories of class libraries much greater in size. Current techniques for managing large class libraries make use of browsers. However, these tend to impose severe limitations when the size of the libraries increases as our experience with developing our own library of collection classes in C++ [1] indicates. This prompted us to look into an alternative approach to tackling the management of large class libraries via a relational data dictionary approach, where data dictionaries are used to describe the information about concepts in C++ and are then implemented using a relational database management system such as INGRES. The proposed technique is capable of providing the developer with an efficient and non-procedural method of retrieving information in a class library.