WCRE11a.ppt
0 likes272 views
This document presents an exploratory study of macro co-changes in software projects. The study examines changes to source code files over time to identify files that frequently change together, even if not always changed by the same developer or in the same commit. It introduces the concept of "macro co-changes" to capture files changed together across different change periods. The study also presents an approach called Macocha that mines version control systems to detect change periods, construct file change profiles, analyze file stability, and identify files with similar macro co-change patterns.
WCRE11a.ppt
- 1. An Exploratory Study of Macro Co-changes Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Hamel, and Giuliano Antoniol Introduction and An Exploratory Study of Macro Co-changes context Problems and Motivation Macocha Fehmi Jaafar, Yann-Ga¨l Gu´h´neuc, Sylvie Hamel, and e e e Empirical study Giuliano Antoniol Validation Universit´ de Montr´al, Qu´bec, Canada e e e Conclusion and Ongoing Work Thursday, October 20, 2011 Pattern Trace Identification, Detection, and Enhancement in Java SOftware Cost-effective Change and Evolution Research Lab
- 2. An Exploratory Study of Macro Introduction Co-changes Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Hamel, and Giuliano Antoniol Introduction and Context context ◮ Developers must continually change their software Problems and Motivation programs to meet new requirements and user needs. Macocha ◮ Many approaches extract and analyse the changes Empirical study undergone by artefacts and infer change propagation.a Validation Conclusion and ◮ Several of these approaches identify co-changes among Ongoing Work artefacts.b a A. E. Hassan and R. C. Holt. ICSM 2004. b Z. Xing and E. Stroulia. TSE 2005. 2 / 23
- 3. An Exploratory Study of Macro Introduction Co-changes Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Hamel, and Co-change Giuliano Antoniol ◮ Two artefacts are co-changing if they were changed by Introduction and the same author and with the same log message in a context Problems and time-window of less than 200 ms.a Motivation ◮ Mockusb defined the proximity in time of check-ins as Macocha the check-in time of adjacent files that differ by less Empirical study than three minutes. Validation Conclusion and ◮ Other studiesc described issues about identifying atomic Ongoing Work change sets and reported that, in all cases, they differed by few minutes. a T. Zimmermann et al. ICSE 2004. b A. Mockus et al. TSE 2004. c D. M. German. TSE 2006. 3 / 23
- 4. An Exploratory Study of Macro Problems and Motivation Co-changes Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Hamel, and Giuliano Antoniol Missing Dependencies Introduction and ◮ If files (e.g., in ArgoUML, context Problems and NotationUtilityJava.java and Motivation ModelElementNameNotationUml.java) were never Macocha changed by the same developer at the same time but Empirical study were changed by developers (mvw and tfmorris) in two Validation consecutive change periods. Conclusion and Ongoing Work ◮ Previous co-changes are intrinsically limited in time. They cannot express patterns of changes between long time intervals (e.g., ArgoDiagram.java and ModeCreateAssociationClass.java were maintained by the same developer but separated by few hours). 4 / 23
- 5. An Exploratory Study of Macro Problems and Motivation Co-changes Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Hamel, and Macro co-change Giuliano Antoniol A macro co-change is two or more files that change together, Introduction and i.e., they were maintained in the same change periods. context Problems and Motivation Macocha Empirical study Validation Conclusion and Ongoing Work Figure: Two changes performed by one developer are sequential in time (after few hours), F1 and F2 are macro co-changing 5 / 23
- 6. An Exploratory Study of Macro Problems and Motivation Co-changes Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Hamel, and Giuliano Antoniol Dephase macro co-change Introduction and A dephase macro co-change is two or more files that have context been observed to change with the same shift s. Problems and Motivation Macocha Empirical study Validation Conclusion and Ongoing Work Figure: Files F1 and F2 are changed by different developers in two 6 / 23 consecutive periods of time. They are dephase macro co-changing
- 7. An Exploratory Study of Macro Macocha Co-changes (1/9) Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Hamel, and Giuliano Antoniol Introduction and context Macocha Problems and Motivation We propose Macocha to: Macocha 1. Mine version-control systems (CVS and SVN). Empirical study 2. Identify the change periods in a program. Validation Conclusion and 3. Group the program source files according to their Ongoing Work stability through the change periods. 4. Identify among changed files those that have similar co-changes pattern. 7 / 23
- 8. An Exploratory Study of Macro Macocha Co-changes (2/9) Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Hamel, and Giuliano Antoniol Introduction and context Macocha Problems and We draw inspiration and extend the classical sliding window Motivation approacha to consider that two subsequent changes are part Macocha of one change period if they were committed by: Empirical study Validation ◮ any author; Conclusion and Ongoing Work ◮ with any log message; ◮ without an interrupt between two changes. a T. Zimmermann et al. ICSE 2004. 8 / 23
- 9. An Exploratory Study of Macro Macocha Co-changes (3/9) Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Hamel, and Macocha Giuliano Antoniol ◮ The duration of a change period is less than 40 hours.a Introduction and context ◮ If the interrupt between two subsequent changes is Problems and more than t = 5.17 hours, theses two changes belong to Motivation two different change periods. Macocha Empirical study ◮ A SMCC is two or more files that have identical profiles Validation during the life cycle of a program. Conclusion and Ongoing Work ◮ A SDMCC is the set composed of F1 and one or more files, F2...FM, such that F2...FM always macro co-change with the same shift in time s with respect to F1. ◮ We use the Hamming distance DH to measure the amount of differences between two change profiles. a L. Hatton. Computer 2007. 9 / 23
- 10. An Exploratory Study of Macro Macocha Co-changes (4/9) Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Hamel, and Giuliano Antoniol Introduction and context Problems and Motivation Macocha Empirical study Validation Conclusion and Ongoing Work Figure: Analysis-process of Macocha 10 / 23
- 11. An Exploratory Study of Macro Macocha Co-changes (5/9) Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Hamel, and Giuliano Antoniol Macocha Introduction and Step 1: Detection of change periods. context Problems and Motivation Macocha Empirical study Validation Conclusion and Ongoing Work Figure: Analysis of commits and creation of change periods 11 / 23
- 12. An Exploratory Study of Macro Macocha Co-changes (6/9) Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Hamel, and Giuliano Antoniol Macocha Introduction and context Step 2: Creation of files profiles. Problems and Motivation Macocha Empirical study Validation Conclusion and Ongoing Work Figure: From revision control systems to file profiles 12 / 23
- 13. An Exploratory Study of Macro Macocha Co-changes (7/9) Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Hamel, and Giuliano Antoniol Introduction and context Macocha Problems and Step 3: Stability analysis. Motivation Macocha Empirical study Validation Conclusion and Ongoing Work Figure: Profiles showing file stability 13 / 23
- 14. An Exploratory Study of Macro Macocha Co-changes (8/9) Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Hamel, and Giuliano Antoniol Macocha Introduction and Step 4: Detection of macro co-changes. context Problems and Motivation Macocha Empirical study Validation Figure: Files F1 and F2 are in macro co-change Conclusion and Ongoing Work Figure: Three different bit vectors showing approximate macro co-changes (D(F1,F2)=2; D(F1,F3)=4) 14 / 23
- 15. An Exploratory Study of Macro Macocha Co-changes (9/9) Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Hamel, and Giuliano Antoniol Macocha Introduction and context Step 5: Detection of dephase macro co-changes. Problems and Motivation Macocha Empirical study Validation Conclusion and Ongoing Work Figure: Three different bit vectors showing dephase macro co-changes 15 / 23
- 16. An Exploratory Study of Macro Empirical study Co-changes Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Hamel, and Giuliano Antoniol Research questions Introduction and context 1. How does Macocha compare to previous work in term Problems and of precision and recall? Motivation Macocha 2. Are there (approximate) dephase macro co-changes Empirical study among files and what is their usefulness? Validation Conclusion and Objects Ongoing Work We now present the results of our empirical study. We apply Macocha on four different programs: ArgoUML, FreeBSD, SIP, and XalanC, developed with three different programming languages, C, C++, and Java. 16 / 23
- 17. An Exploratory Study of Macro Empirical study Co-changes Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Hamel, and Giuliano Antoniol Objects Introduction and ArgoUML FreeBSD SIP XalanC context Languages Java C Java C++ Problems and Versions 30 8 2 21 Motivation Files 3,148 3,603 2,790 529 Macocha Changes 16,727 186,959 8,046 397,052 Empirical study Start Dates 98-01-26 94-05-25 05-07-21 99-12-18 Validation End Dates 09-01-29 09-02-11 10-12-09 09-01-17 Conclusion and CPs 2,843 1,121 1,553 924 Ongoing Work Table: Descriptive statistics of the object programs (CPs: numbers of change periods) 17 / 23
- 18. An Exploratory Study of Macro Empirical study Co-changes Fehmi Jaafar, Yann-Ga¨l e Results Gu´h´neuc, Sylvie e e Hamel, and ArgoUML FreeBSD SIP XalanC Giuliano Antoniol Idle files 202 1,856 963 7 Changed files 2,946 1,747 1,827 522 Introduction and context # of SMCC 166 121 142 36 Problems and Max # files 35 24 15 17 Motivation Min # files 2 2 2 2 Macocha # of SMCCH 196 163 182 85 Empirical study Max # files 46 44 32 22 Validation Min # files 2 2 2 2 Conclusion and # of SDMCC 11 1 6 1 Ongoing Work Max # files 4 2 3 2 Min # files 2 2 2 2 # of SDMCCH 53 63 36 4 Max # files 6 8 5 2 Min # files 2 2 2 2 Table: Cardinalities of the sets obtained in the empirical study 18 / 23
- 19. An Exploratory Study of Macro Validation Co-changes (1/3) Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Hamel, and Giuliano Antoniol Validation Introduction and context We perform two types of validation: Problems and Motivation ◮ Quantitatively, we compare the stability analysis of Macocha Macocha with that of UMLDiffa and the co-change Empirical study analysis of Macocha with association rulesb . Validation ◮ Qualitatively, we use external information provided by Conclusion and Ongoing Work bugs reports, mailing lists, and requirement descriptions to validate the (dephase) macro co-changes not found using association rules. a Z. Xing et al. ICSE 2005. b T. Zimmermann et al. ICSE 2004. 19 / 23
- 20. An Exploratory Study of Macro Validation Co-changes (2/3) Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Hamel, and Giuliano Antoniol Validation Introduction and context Idle Groups Changed Groups Problems and Idle Clusters 202 0 Motivation ArgoUML Short-lived Clusters 0 1,390 Macocha Active Clusters 0 1,556 Empirical study Idle Clusters 963 0 Validation SIP Short-lived Clusters 0 997 Conclusion and Active Clusters 0 830 Ongoing Work Idle Clusters 7 0 XalanC Short-lived Clusters 0 291 Active Clusters 0 231 Table: Cardinality of Macocha sets in comparison to UMLDiff 20 / 23
- 21. An Exploratory Study of Macro Validation Co-changes (3/3) Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Validation Hamel, and Giuliano Antoniol Association Rules Macocha Precision Recall Precision Recall Introduction and context ArgoUML 15% 66% 20% 75% Problems and FreeBSD 22% 100% 24% 100% Motivation SIP 18% 89% 24% 91% Macocha XalanC 16% 100% 22% 100% Empirical study Table: Association rules’s approach vs. Macocha Validation Conclusion and Association Rules External Information Ongoing Work Precision Recall Precision Recall ArgoUML 86% 98% 100% 99% FreeBSD 98% 100% 100% 100% SIP 85% 96% 100% 98% XalanC 90% 100% 100% 100% Table: External evaluation of Macocha when using the results of the association rules’s approach as oracle and after manual validation using external information 21 / 23
- 22. An Exploratory Study of Macro Conclusion and Ongoing Work Co-changes Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Hamel, and Giuliano Antoniol Conclusion Introduction and context 1. We introduced the novel concepts of macro co-changes Problems and and dephase macro co-changes to describe that two files Motivation were changed by developers within same change Macocha periods, with possible shifts in time. Empirical study Validation 2. We described, Macocha, an approach to detect Conclusion and (dephase) macro co-changes using file profiles and their Ongoing Work stability in time. 3. We performed two types of validations: quantitatively and qualitatively, and we showed that SMCC and SDMCC do exist and bring supplementary information. 22 / 23
- 23. An Exploratory Study of Macro Conclusion and Ongoing Work Co-changes Fehmi Jaafar, Yann-Ga¨l e Gu´h´neuc, Sylvie e e Hamel, and Giuliano Antoniol Introduction and Ongoing Work context Problems and 1. Performing a comprehensive study of the number of Motivation MCCs and DMCCs with varying values of t and s. Macocha 2. Performing a comprehensive study of the different kinds Empirical study of DMCCs. Validation Conclusion and 3. Relating MCCs and DMCCs with static analysis and Ongoing Work external software characteristics, such as change proneness. 23 / 23