Evaluating design decay during software evolution

Software systems evolve, requiring continuous maintenance and development. They undergo changes throughout their lifetimes as new features are added and bugs are fixed. As these systems evolved, their designs tend to decay with time and become less adaptable to changing users’ requirements. Consequently, software designs become more complex over time and harder to maintain; in some not-so-rare cases, developers prefer redesigning from scratch rather than prolonging the life of existing designs, which causes development and maintenance costs to rise. Therefore, developers must understand the factors that drive the decay of their designs and take proactive steps that facilitate future changes and slow down decay. Design decay occurs when changes are made on a software system by developers who do not understand its original design. On the one hand, making software changes without understanding their effects may lead to the introduction of bugs and the premature retirement of the system. On the other hand, when developers lack knowledge and–or experience in solving a design problem, they may introduce design defects, which are conjectured to have a negative impact on the evolution of systems, which leads to design decay. Thus, developers need mechanisms to understand how a change to a system will impact the rest of the system and tools to detect design defects. In this dissertation, we propose three principal contributions. The first contribution aims to evaluate design decay. Measuring design decay consists of using a diagram matching technique to identify structural changes among versions of a design, such as a class diagram. Finding structural changes occurring in long-lived, evolving designs requires the identification of class renamings. Thus, the first step of our approach concerns the identification of class renamings in evolving designs. Then, the second step requires to match several versions of an evolving design to identify decaying and stable parts of the design. We propose bit-vector and incremental clustering algorithms to match several versions of an evolving design. The third step consists of measuring design decay. We propose a set of metrics to evaluate this design decay. The second contribution is related to change impact analysis. We present a new metaphor inspired from seismology to identify the change impact. In particular, our approach considers changes to a class as an earthquake that propagates through a long chain of intermediary classes. Our approach combines static dependencies between classes and historical co-change relations to measure the scope of change propagation in a system, i.e., how far a change propagation will proceed from a “changed class” to other classes. The third contribution concerns design defect detection. We propose a metaphor inspired from a natural immune system. Like any living creature, designs are subject to diseases, which are design defects. Detection approaches are defense mechanisms of designs. A natural immune system can detect similar pathogens with good precision. This good precision has inspired a family of classification algorithms, Artificial Immune Systems (AIS) algorithms, which we use to detect design defects. The three contributions are evaluated on open-source object-oriented systems and the obtained results enable us to draw the following conclusions: • Design decay metrics, Tunnel Triplets Metric (TTM) and Common Triplets Metric ( CTM), provide developers useful insights regarding design decay. If TTM decreases, then the original design decays. If TTM is stable, then the original design is stable, which means that the system is more adapted to the new changing requirements. • Seismology provides an interesting metaphor for change impact analysis. Changes propagate in systems, like earthquakes. The change impact is most severe near the changed class and drops off away from the changed class. Using external information, we show that our approach helps developers to locate easily the change impact. • Immune system provides an interesting metaphor for detecting design defects. The results of the experiments showed that the precision and recall of our approach are comparable or superior to that of previous approaches.

Récupéré en direct depuis OpenAlex et désinversé. Les résumés ne sont pas conservés dans cette base de données : les index inversés représentent 8,6 Go des 9,3 Go de texte de la base, et le serveur dispose de 13 Go libres.