RefDiff: Detecting Refactorings in Version Histories (MSR 2017)

Refactoring is a well-known technique that is widely adopted by software engineers to improve the design and enable the evolution of a system. Knowing which refactoring operations were applied in a code change is a valuable information to understand software evolution, adapt software components, merge code changes, and other applications. In this paper, we present RefDiff, an automated approach that identifies refactorings performed between two code revisions in a git repository. RefDiff employs a combination of heuristics based on static analysis and code similarity to detect 13 well-known refactoring types. In an evaluation using an oracle of 448 known refactoring operations, distributed across seven Java projects, our approach achieved precision of 100% and recall of 88%. Moreover, our evaluation suggests that RefDiff has superior precision and recall than existing state-of-the-art approaches.

Transcript

1. RefDiff: Detecting Refactorings in Version Histories Danilo Silva, Marco Tulio Valente Universidade Federal de Minas Gerais Belo Horizonte, Brazil
2. Introduction • Software components are in constant change • One important kind of change is refactoring 2
3. Introduction • Knowledge of the refactoring operations applied is a valuable information – Analyze software evolution – Study refactoring practice – Review and merge code 3
4. Problem: Finding refactoring activity is a non- trivial task 4
5. Problem: Finding refactoring activity is a non- trivial task Documentation? • Refactorings are rarely documented 5
6. Problem: Finding refactoring activity is a non- trivial task Instrumenting refactoring engines? • Refactorings are not always performed using automated support 6
7. Problem: Finding refactoring activity is a non- trivial task Source code analysis? • Viable, but current approaches have precision and recall issues – Refactoring Miner: 63% precision – Ref-Finder: 35% precision and 24% recall 7
8. PROPOSED SOLUTION 8
9. RefDiff 9 • A refactoring detection approach – Employs a combination of heuristics based on static analysis and code similarity – 13 well-known refactoring types – TF-IDF based similarity index
10. RefDiff: Overview 10 Version before Version after Input
11. RefDiff: Overview 11 Version before Version after Source Code Analysis Types, methods, and fields }
12. RefDiff: Overview 12 Version before Version after Relationship Analysis Rename Extract Move
13. Relationships 13
14. Relationship Example: Rename Method 14
15. Relationship Example: Rename Method 15 names of mb and ma should be different
16. Relationship Example: Rename Method 16 mb and ma are in “the same” class
17. Relationship Example: Rename Method 17 the similarity index between mb and ma should be greater than a threshold
18. Computing Similarity 18 • Source code represented as a multiset (or bag) of tokens • Similarity index based on Information Retrieval techniques (TF-IDF)
19. Calibration of Thresholds 19 • Oracle of known refactorings in 10 commits of a public dataset (Silva et al., 2016) • Thresholds from 0.1 to 0.9 by 0.1 increments • We choose the value that optimize the F1 score
20. Calibration Results 20
21. EVALUATION 21
22. Evaluation: Precision and Recall 22 • Oracle of known refactorings applied by students – 7 open-source systems – 448 refactoring relationships • Compare RefDiff’s precison and recall with – Refactoring Miner – Refactoring Crawler – Ref-Finder
23. Evaluation: Precision and Recall 23
24. Conclusion 24 • RefDiff has better precision and recall than other approaches • Execution time is acceptable (1.96s per commit)
25. Future Work 25 • Extended evaluation of RefDiff using actual refactorings applied in open-source systems
26. THANK YOU! https://github.com/aserg-ufmg/RefDiff
27. Evaluation: Precision and Recall 27
28. Evaluation: Execution Time 28
29. Evaluation: Execution Time 29 • We analyzed each commit between January 1, 2017 and March 27, of 10 Java repositories – 1990 commits • We compared execution time with Refactoring Miner
30. Evaluation: Execution Time 30 Approach Avg. time (s) Total time(s) RefDiff 1.96 3,893 Ref. Miner 0.89 1,779
31. Computing Similarity: Example 31
32. Computing Similarity: Example 32
33. Computing Similarity: Example 33
34. Computing Similarity 34 token frequency in entity e inverse document frequency of the token in the collection weighted Jaccard coefficient