Quality of the Source Code for Design and Architecture Recovery Techniques: Utilities are the Problem

Software maintenance is perhaps one of the most difficult activities in software engineering, especially for systems that have undergone several years of ad hoc maintenance. The problem is that, for such systems, the gap between the system implementation and its design models tend to be considerably large. Reverse engineering techniques, particularly the ones that focus on design and architecture recovery, aim to reduce this gap by recovering high-level design views from the source code. The course code becomes then the data on which these techniques operate. In this paper, we argue that the quality of a design and architecture recovery approach depends significantly on the ability to detect and eliminate the unwanted noise in the source code. We characterize this noise as being the system utility components that tend to encumber the system structure and hinder the ability to effectively recover adequate design views of the system. We support our argument by presenting various design and architecture recovery studies that have been shown to be successful because of their ability to filter out utility components. We also present existing automatic utility detection techniques along with the challenges that remain unaddressed.