Reverse engineering: A cognitive approach, a case study and a tool.

Abstract

Software maintenance (SM) for large legacy systems is a very inefficient process; on average, 70% of software costs are spent on maintenance [Swanson 89]. The inefficiency of SM has been related to the difficulty comprehending software systems; therefore program comprehension is considered to be a key bottleneck of SM. Reverse engineering tools have been used to alleviate this bottleneck with lower than expected success. We present a cognitively based approach for reverse engineering tool development. We use ideas from cognitive psychology and other disciplines to formulate the approach. We also describe a case study in which we applied the approach in a telecommunication company. The case study resulted in the development of DynaSee, a reverse engineering tool which helps software engineers analyze dynamic program information. DynaSee reads routine call traces, and provides several processing and visualization features that make the use of traces much more useful for software maintenance and program comprehension. Next, we describe and evaluate the various features of DynaSee that compress, abstract and augment traces to make them comprehensible and useful for SM tasks. Finally, based on our experience in developing DynaSee, we generalize the aspects of our findings and techniques that are based on psychology by relating them to the mainstream psychological literature and to other disciplines where similar techniques have been used.