An Empirical Study on Recommendations of Similar Bugs (SANER 2016)

The work to be performed on open source systems, whether feature developments or defects, is typically described as an issue (or bug). Developers self-select bugs from the many open bugs in a repository when they wish to perform work on the system. This paper evaluates a recommender, called NextBug, that considers the textual similarity of bug descriptions to predict bugs that require handling of similar code fragments. First, we evaluate this recommender using 69 projects in the Mozilla ecosystem. We show that for detecting similar bugs, a technique that considers just the bug components and short descriptions perform just as well as a more complex technique that considers other features. Second, we report a field study where we monitored the bugs fixed for Mozilla during a week. We sent mails to the developers who fixed these bugs, asking whether they would consider working on the recommendations provided by NextBug; 39 developers (59%) stated that they would consider working on these recommendations; 44 developers (67%) also expressed interest in seeing the recommendations in their bug tracking system

Transcript

1. An Empirical Study on Recommendations of Similar Bugs Henrique Rocha (UFMG, Brazil) Marco Túlio Valente (UFMG, Brazil) Humberto Marques-Neto (PUC-Minas, Brazil) Gail Murphy (UBC, Canada) SANER 2016
2. Introduction n  Evaluate the feasibility of recommending similar bugs to open source developers o  More efficient to work on similar code o  Less effort to locate and understand code 2
3. Motivation 3
4. Challenge n  Open source systems have many issues (or bugs) and developers o  Mozilla 2.8K fixed issues per month and 2K contributors (2012). n  Can we predict similar bugs? o  Bugs with similar code fragments o  Using only bug report fields 4
5. Goal n  Evaluate 2 techniques for recommending similar bugs o  NextBug [Rocha et al. JSERD 2015] o  REP [Sun et al. ASE 2011] 5
6. Evaluated Techniques 6
7. NextBug 7
8. NextBug 8
9. REP n  Proposed to detect duplicate bug reports n  Uses 7 features to compute bug similarity o  Textual description (unigram & bigram), component, product, version, type, & priority. n  Two-round parameter training 9
10. Comparative Study 10
11. Study Design n  Dataset: 65K Mozilla bugs (2009 – 2012) n  Metrics: Feedback, Precision, Likelihood, Recall, and F-score n  Oracle: Two bugs are similar if the set of changed files are at least 50% similar 11
12. Feedback 12
13. Precision 13
14. Execution Time (ms) 14
15. Field Study 15
16. Survey Setup n  We showed NextBug recommendations for real bugs to 176 Mozilla developers o  66 devs (37%) responded o  We asked two survey questions 16
17. Question #1 Would you consider working on one of the recommended bugs? 17 “The suggestions seem pretty accurate (...) the #3 bug is new to me — I didn’t know about that one” (Subject #47) “Bugs may interest me but are definitely not my current focus” (Subject #1)
18. Question #2 Would you consider useful a Bugzilla extension with recommendations? 18 “(...) would be immensely helpful for new contributors as they don’t know the project very well” (Subject #10) “I don’t manage a lot of open bugs at the same time, so I don’t need a list of suggestions” (Subject #22)
19. Final Remarks 19
20. Conclusions n  Comparative study (65K bugs) o  NextBug & REP show similar results o  NextBug has a better execution time n  Field Study: Devs confirmed that NextBug can provide useful recommendations 20
21. henrique.rocha@dcc.ufmg.br mtov@dcc.ufmg.br humberto@pucminas.br murphy@cs.ubc.ca Thank you! Arigatou gozaimasu! Questions? SANER 2016