Architecture Recovery by Semi-Automatic Component Identification

AbstractArchitecture recovery is the process of analyzing a system in order to identify the system's components, and their connectors. This paper gives a presentation of dominance analysis, and how it can be used to identify software components in object oriented legacy systems. The actual dominance analysis is applied on a high level representation of the system that we refer to as the class graph, a directed graph where the nodes are the system classes, and the edges correspond to classes interacting with each other. The result indicates that dominance analysis is a useful tool to identify certain types of components (the passive components) but insufficient to recover the complete system architecture. We also discuss how dominance analysis can be combined with other identification methods to recover a larger part of the architecture, and a novel approach to merge the results from different identification methods into a single data structure, a semi-lattice over the power set of all system classes, that permits a unified view of the different results