This paper presents results of a research programme on reverse engineering using a transformation system for maintenance and focuses on dealing with data-intensive programs such as those written in COBOL. Problems with data-intensive programs are addressed, our solutions to these problems are discussed and the results of experiments are presented. It is concluded that formal transformations provide a way of combining design decisions which have become instantiated in both the code and the data structures.
We describe a solution to the extraction of Entity Relationship (ER) diagrams from data-intensive source code. In such programs, the relationships between data items are often represented within imperative code as well as within data structures, and we show that reverse engineering can be improved if both are used. This distinguishes our work from other works in the field.
Our method is based on formal transformations. We identify imperative constructs which improve the high-level ER models that can be captured. Suitable transformations are then briefly summarised. A series of experiments with industrial COBOL programs is described. Our results show that code-embedded relations can be usefully incorporated into data intensive reverse engineering, and that they enhance the designs extracted.