The Need for Inner-Procedural Refinements

Today one of the most widely accepted paradigms of software engineering is the Top-Down method. Based on the hypothesises that (a) Top-Down is a good way to go and (b) programmers are lazy by their very nature, I show here that with most modern imperative programming languages, we stop the Top-Down process at least one level too early. What languages need is the concept of refinements within procedures, so the Top-Down development process can continue arbitrarily deep inside each procedure. This leads to better-structured code with improved understandability. Existing languages can be extended to support the notion of inner-procedural refinements. It is described how this was done with a fast preprocessor for C and C ++ and what the experiences with it are. Key words: refinements, C, programming language extension, Top-Down method The Top-Down design process Let us take a quick look on why Top-Down is a good method. We will see that the ultimate reason lies in the human nature. Toda..