Towards a Computer Understanding of Program Design

Abstract

Program design is one of the many processes involved in program development and is considered to be essential to the development of structured programs. Consequently, this research has been concerned with the analysis of program design since it is considered to be of equal importance to other areas of Artificial Intelligence (AI) research, which analyse the program code. Because a rigorous program design results in a program containing few errors, a system capable of analysing program designs should assist these other related areas of AI. This research has developed the Framework for Analysing Program Designs (or FAPD) in order to analyse the kinds of program design produced by programmers using the principles of structured programming. The process of analysis is viewed as comprising four distinct phases, which are referred to as pre-semantic analysis, semantic analysis, generation of comments, and code generation. The results of analysis take the form of a coded version of the program design together with any comments about the code. Analysis is based on a set of structures which have been developed in order to represent phrases and statements often used in a program design. Attached to each structure is a procedure, referred to as a class instance, which translates its structure into a particular programming language. FAPD has been implemented and tested within a system called DACE (which is a Design Analysing and Commenting Environment). FAPD is discussed within the context of the system and the results from testing it are discussed in detail. The conclusions are drawn that FAPD represents a viable approach to the computer analysis of program designs, the system has some influence on those who use it and that class instances are a useful acquisition to the set of tools currently available to researchers in AI.