Analyzing the evolutionary history of the logical design of object-oriented software

• Published in: IEEE transactions on software engineering Vol. 31; no. 10; pp. 850 - 868
• Main Authors: Xing, Z., Stroulia, E.
• Format: Journal Article
• Language: English
• Published: New York IEEE 01.10.2005; IEEE Computer Society
• Subjects: Algorithm design and analysis; Algorithms; and reengineering; Case studies; Computer programs; Design engineering; Documentation; Evolution; Evolutionary; History; Index Terms- Restructuring; Mathematical models; Object oriented; Object oriented modeling; Object oriented programming; Project management; Quality management; Reverse engineering; Software; Software design; Software engineering; Software systems; Studies; Surges; Systems design; Visualization
• ISSN: 0098-5589; 1939-3520
• DOI: 10.1109/TSE.2005.106

Today, most object-oriented software systems are developed using an evolutionary process model. Therefore, understanding the phases that the system's logical design has gone through and the style of their evolution can provide valuable insights in support of consistently maintaining and evolving the system, without compromising the integrity and stability of its architecture. In this paper, we present a method for analyzing the evolution of object-oriented software systems from the point of view of their logical design. This method relies on UMLDiff, a UML-structure differencing algorithm, which, given a sequence of UML class models corresponding to the logical design of a sequence of system code releases, produces a sequence of "change records" that describe the design-level changes between subsequent system releases. This change-records sequence is subsequently analyzed from the perspective of each individual system class, to produce the class-evolution profile, i.e., a class-specific change-records' sequence. Three types of longitudinal analyses - phasic, gamma, and optimal matching analysis - are applied to the class-evolution profiles to recover a high-level abstraction of distinct evolutionary phases and their corresponding styles and to identify class clusters with similar evolution trajectories. The recovered knowledge facilitates the overall understanding of system evolution and the planning of future maintenance activities. We report on one real-world case study evaluating our approach.