Learning and Suggesting Source Code Changes from Version History: A Systematic Review

Context: Software systems are in continuous evolution through source code changes to fixing bugs, adding new functionalities and improving the internal architecture. All these practices are recorded in the version history, which can be reused as an advantage in the development process. Objective: This paper aims to investigate approaches and techniques related to the learning of source code changes, since the change identification step, learning, and reuse in recommending strategies. Method: We conducted a systematic review related to primary studies about source code changes. The search approach identified 2410 studies, up to and including 2012, which resulted in a final set of 39 selected papers. We grouped the studies according to each established research question. This review investigates how source code changes, which were performed in the past of software, can support the improvement of the software project. Results: The majority of approaches and techniques have used repetitiveness behavior of source code changes to identify structural or metrics patterns in software repositories, trough the evaluation of sequences of versions. To extract the structural patterns, the approaches have used programming-by-example techniques to differencing source code changes. In quality metrics analysis, the studies have applied mainly complexity and object-oriented metrics. Conclusion: The main implication of this review is that source code changes as examples, to support the improvement of coding practice during the development process, in which we presented some relevant strategies to guide each step, since identifying until the suggesting of source code changes.