The use of non-formal information in reverse engineering and software reuse

Abstract

Within the field of software maintenance, both reverse engineering and software reuse have been suggested as ways of salvaging some of the investment made in software that is now out of date. One goal that is shared by both reverse engineering and reuse is a desire to be able to redescribe source code, that is to produce higher level descriptions of existing code. The fundamental theme of this thesis is that from a maintenance perspective, source code should be considered primarily as a text. This emphasizes its role as a medium for communication between humans rather than as a medium for human-computer communication. Characteristic of this view is the need to incorporate the analysis of non-formal information, such as comments and identifier names, when developing tools to redescribe code. Many existing tools fail to do this. To justify this text-based view of source code, an investigation into the possible use of non-formal information to index pieces of source code was undertaken. This involved attempting to assign descriptors that represent the code's function to pieces of source code from IBM's CICS project. The results of this investigation support the view that the use of nonformal information can be of practical value in redescribing source code. However, the results fail to suggest that using non-formal information will overcome any of the major difficulties associated with developing tools to redescribe code. This is used to suggest future directions for research.