IntelliMerge: a refactoring-aware software merging technique

Abstract

IntelliMerge: A Refactoring-Aware Software Merging Technique 170:3 generality and efficiency of automatic conflicts resolving by combining unstructured and structured approaches. These approaches can also be classified as textual, syntactic and semantic approaches, according to the extent they make use of the semantic information in the source code [Mens 2002]. However, most of the existing approaches still fail to deal with one kind of complex changes made to source code, namely, the refactorings. In general, a refactoring change to a program denotes a transformation to the program (e.g., Rename/Move Field and Extract/Inline Method) that improves the program's internal design without changing its externally observable behavior [Fowler 2002]. Complex refactorings usually consist of a set of related changes across multiple source code files in the program. As an important kind of change in object-oriented programming (OOP) and agile development, refactorings are widespread in software development and evolution [Fowler 2002]. The automated refactoring tools also contribute to the popularity of refactorings since these tools allow programmers to apply refactoring quickly and safely [Dig et al. 2007]. However, it has been observed that refactoring is one of the main factors that make trouble to unstructured merge techniques [Dig et al. 2006a, 2007]. A recent empirical study [Mahmoudi et al. 2019] on about 3,000 open-source Java projects also concludes that (1) at least 22% of merge conflicts are related to refactoring, and (2) these conflicts are more complex and difficult to resolve comparing with those refactoring-irrelevant conflicts. On one hand, a refactoring often affect multiple parts of the program, but most existing approaches do not consider the possible dependencies between different changes in the merging process; on the other hand, refactoring-related conflicts are more difficult for developers to understand and resolve, because of the absence of the refactoring history [Tsantalis et al. 2018].

Towards a Catalog of Composite Refactorings

“…When used in a context like this one, a refactoring detection tool, such as RefDiff or RefactoringMiner [5][6][7][8][9][10], detects the six single refactorings independently. However, it would also be interesting to detect a high-level Techniques that may benefit from the detection of independent refactorings (like code visualization [12][13][14], code review [15][16][17][18][19], code authorship [20,21], bug-introducing detection [22,23], refactoring-aware tools [24,25], software mining approaches [26][27][28][29], to name a few) may also benefit from the detection of composite refactorings. As refactoring detection is the basis for such techniques, composite refactorings would bring to light novel operations not restricted to time and scope, therefore, better representing the actual source code changes.…”

RAID: Tool Support for Refactoring-Aware Code Reviews

“…For instance, there are studies on motivations to perform refactoring operations [6], [12]- [14], challenges and benefits [15], [16], the impact on software quality [17]- [20], security [21]- [23], and also on tools to assist developers during refactoring tasks [24]- [30]. Besides, various studies focused on the identification of refactoring operations through analysis of version control systems [7]- [11], [25].…”

“…Textual diffs also present issues when merging changes in version control systems. Bo et al [25] proposed a graph-based refactoring-aware algorithm to detect and resolve refactoringrelated conflicts. The algorithm was evaluated with 1,070 merge scenarios from popular open source Java projects and archived a 58.90% reduction of merge conflicts comparing with GitMerge and 11.84% over jFSTMerge.…”

A constructive model for collective intelligence

“…Solutions to this problem are closely related to existing AI research, including knowledge graphs, recommendation algorithms, etc. Currently, we are exploring graph-based approaches to this problem in the fields of software development [ 9 ] and jigsaw puzzles [ 10 ]: the former studies the value of graph-based information representation in program merging; the latter focuses on the resolving of EIFL-based jigsaw puzzles, and can be experienced at https://www.pintu.fun .…”