Analysing large COBOL programs: the extraction of reusable modules

We describe work performed for the IDENT project which is concerned with the identification and encapsulation of reusable modules. The aim of the project is to look at the integration of two methods for supporting maintenance; RE/sup 2/ and RECAST. We describe a case study where we try to extract reusable modules from a 21,000 line COBOL program. We approach this through the use of some of the basic methods addressed by the RE/sup 2/ reuse reengineering paradigm and some of the stages of the reverse engineering RECAST method. We have found that the identification of modules is very difficult because of high connectivity between data items. In order to deal with this problem we identify a number of strategies for removing particular data items using some of the RECAST stages and steps. These include the logical data model and system processing steps. However, in most cases we have found that such approaches were insufficient and that expert domain knowledge is also required to assist in module identification.