Recovering the Structure of Software Systems Using Tube Graph Interconnection Clustering

An important product of the software design phase is the speciication of the software structure at various levels of detail. Without reliable design documentation, signiicant software systems become less accessible to software engineers because structural information is buried in the intricate implementation source code. Reverse engineering techniques aim at recovering the structure of software systems, from the source code and mental models of developers, in order to make these systems more understandable to those maintaining them. Many reverse engineering techniques rely on creating a decomposition hierarchy by recursively clustering related software components (e.g., variables, procedures, classes, modules) into composite components (e.g., subsystems). Component clustering is necessary for managing complexity, and therefore is an important step in the reverse engineering process. In this paper, we argue that the clustering of interconnections is also necessary during reverse engineering. We propose an approach, based on a formalism called tube graphs, to specifying system structure that considers both composite software components and composite inter-connections between these components. We present eecient algorithms for software inter-connection clustering, which nicely complement existing software component clustering algorithms.