Recovery of Object-Oriented Design Patterns Using Static and Dynamic

Abstract Maintenance is undoubtedly an effort-consuming activity in the software process. The discovery of a collection of high level building blocks, such as design patterns, in a software artifact represents an important step in program understanding and, hence, can be extremely useful during maintenance. This paper presents a new approach that combines static and dynamic analyses to improve the precision and efficiency of recovering object-oriented design patterns. First, we recover pattern candidates, which adhere to the static structure of different patterns. The static recovery algorithms utilize new abstractions, called basic units, to improve the static identification efficiency. Then, two static constraints are enforced on the recovered candidates to improve the identification precision. Finally, we apply dynamic analysis through profiling to further improve the precision. For each pattern, only relevant methods, which characterize its dynamic behaviour, are traced. The overall analysis approach is automated and evaluated on several large-scale open-source case studies. Results of the case studies and the comparisons demonstrate the improved precision and efficiency of the presented approach.