Structure-based clustering of components for software reuse

The characterization of the code reuse practices in existing production environments provides fundamental data and lessons for the establishment or improvement of effective reuse-oriented policies, and for the adoption of up-to-date technologies supporting them. The method and results of an experience of metric-aided clustering of software components, aimed at detecting and characterizing implicit reuse of code and reuse potential in a large-scale data processing environment, are presented. Similar function may be in fact replicated many times, customizing an existing source code component, but this phenomenon may be only partially apparent in the form of explicit reuse. A set of software metrics has been used to create clusters of existing components whose internal structures appear very similar. Functional similarity checks involving human experts were then performed. This was done in the context of a large reuse project, where quantitative software quality indicators are also combined with the feedback collected in pilot groups who know the applications from which the candidate components were extracted. The potential and limitations of metric support in this field are considered in the discussion of the results obtained