Detecting Design Rule Violations

In this paper, we present an approach to detect design rule violations that could cause software defects, modularity decay, or expensive refactorings. Our approach is to compute the discrepancies between how components should change together based on the modular structure framed by design rules, and how components actually changed together revealed by how modification requests were fulfilled. Our contributions include a design violation detection framework and a design-rule based impact scope prediction algorithm. We evaluated our approach using the version history of three large-scale open source software projects. We examined all identified violations to check whether they were refactored or recognized by the developers in later versions. Our results show that (1) on average 73% of the violations we identified were either recognized or refactored in later releases (when using .5 confidence and varying support from 2 to 10 in Hadoop); (2) our approach can identify problematic design violations much earlier than manual identification by developers; and (3) the identified violations cover multiple bad smells, such as tight coupling and code clone.