ICSM10a.ppt
0 likes237 views
This document discusses measuring and relating identifier dispersion to fault proneness. It introduces physical and conceptual dispersion measures, including entropy and context coverage. An aggregated metric called numHEHCC is presented that counts the number of terms with high entropy and different contexts within an entity. The study aims to examine if numHEHCC captures characteristics beyond size alone and to determine if term entropy and context coverage help explain faults in an entity.
![Physical and
Conceptual Introduction
Identifier
Dispersion
Venera
Arnaoudova, Laleh
Eshkevari, Rocco
Oliveto, Yann-Ga¨l
e
Gu´h´neuc,
e e Fault identification
Giuliano Antoniol
size (e.g., [Gyim´thy et al., 2005])
o
Introduction
cohesion (e.g., [Liu et al., 2009])
Our study
coupling (e.g., [Marcus et al., 2008])
Dispersion
measures
number of changes (e.g., [Zimmermann et al., 2007])
Our study - refined Importance of linguistic information
Case study
RQ1 – Metric Relevance program comprehension (e.g.,
RQ2 – Relation to Faults
[Takang et al., 1996, Deissenboeck and Pizka, 2006,
Conclusions and
future work
Haiduc and Marcus, 2008, Binkley et al., 2009])
code quality (e.g., [Marcus et al., 2008,
Poshyvanyk and Marcus, 2006, Butler et al., 2009])
3 / 16](https://image.slidesharecdn.com/icsm10a-ppt-130412083605-phpapp01/85/ICSM10a-ppt-3-320.jpg)
ICSM10a.ppt
- 1. Physical and Conceptual Identifier Dispersion Venera Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Physical and Conceptual Identifier Dispersion: Gu´h´neuc, e e Giuliano Antoniol Measures and Relation to Fault Proneness Introduction Our study Dispersion Venera Arnaoudova Laleh Eshkevari Rocco Oliveto measures Yann-Ga¨l Gu´h´neuc Giuliano Antoniol e e e Our study - refined Case study ´ SOCCER Lab. – DGIGL, Ecole Polytechnique de Montr´al, Qc, Canada e RQ1 – Metric Relevance SE@SA Lab – DMI, University of Salerno - Salerno - Italy RQ2 – Relation to Faults ´ Ptidej Team – DGIGL, Ecole Polytechnique de Montr´al, Qc, Canada e Conclusions and future work September 15, 2010 SOftware Cost-effective Change and Evolution Research Lab Software Engineering @ SAlerno Pattern Trace Identification, Detection, and Enhancement in Java
- 2. Physical and Conceptual Outline Identifier Dispersion Venera Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Introduction Gu´h´neuc, e e Giuliano Antoniol Our study Introduction Our study Dispersion measures Dispersion measures Our study - refined Our study - refined Case study RQ1 – Metric Relevance RQ2 – Relation to Faults Case study Conclusions and future work RQ1 – Metric Relevance RQ2 – Relation to Faults Conclusions and future work 2 / 16
- 3. Physical and Conceptual Introduction Identifier Dispersion Venera Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Gu´h´neuc, e e Fault identification Giuliano Antoniol size (e.g., [Gyim´thy et al., 2005]) o Introduction cohesion (e.g., [Liu et al., 2009]) Our study coupling (e.g., [Marcus et al., 2008]) Dispersion measures number of changes (e.g., [Zimmermann et al., 2007]) Our study - refined Importance of linguistic information Case study RQ1 – Metric Relevance program comprehension (e.g., RQ2 – Relation to Faults [Takang et al., 1996, Deissenboeck and Pizka, 2006, Conclusions and future work Haiduc and Marcus, 2008, Binkley et al., 2009]) code quality (e.g., [Marcus et al., 2008, Poshyvanyk and Marcus, 2006, Butler et al., 2009]) 3 / 16
- 4. Physical and Conceptual Our study Identifier Dispersion Venera Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l Gu´h´neuc, e e e Term dispersion Giuliano Antoniol We are interested in studying the relation between term Introduction dispersion and the quality of the source code. Our study Dispersion term basic component of identifiers measures dispersion the way terms are scattered among different Our study - refined entities (attributes and methods) Case study RQ1 – Metric Relevance quality absence of faults RQ2 – Relation to Faults Conclusions and future work Example: What is the impact of using getRelativePath, returnAbsolutePath, and setPath as method names on the fault proneness of those methods? 4 / 16
- 5. Physical and Conceptual Dispersion measures Identifier Dispersion (1/3) Venera Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l Gu´h´neuc, e e e Physical dispersion - Entropy Giuliano Antoniol Terms Introduction Entropy Our study Dispersion fee measures Our study - refined Case study foo RQ1 – Metric Relevance RQ2 – Relation to Faults Conclusions and future work bar Entities E1 E2 E3 E4 E5 The circle indicates the occurrences of a term in an entity. The higher the size of the circle the higher the number of occurrences. 5 / 16
- 6. Physical and Conceptual Dispersion measures Identifier Dispersion (2/3) Venera Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Conceptual dispersion - Context Coverage Gu´h´neuc, e e Giuliano Antoniol Entity Contexts C4 Introduction C1 E4 C2 Our study E1 E2 Dispersion measures E5 Terms Context Our study - refined E3 coverage Case study C3 RQ1 – Metric Relevance fee RQ2 – Relation to Faults Entity contexts are identified taking into account the terms contained in the entities. Conclusions and foo future work bar C1 C2 C3 C4 Contexts The star indicates that the term appears in the particular context. 6 / 16
- 7. Physical and Conceptual Dispersion measures Identifier Dispersion Aggregated metric - numHEHCC Venera (3/3) Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Gu´h´neuc, e e Entropy Giuliano Antoniol Introduction th H Our study Dispersion measures Our study - refined Case study RQ1 – Metric Relevance RQ2 – Relation to Faults Conclusions and future work th CC Context Coverage 7 / 16
- 8. Physical and Conceptual Dispersion measures Identifier Dispersion Aggregated metric - numHEHCC Venera (3/3) Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Gu´h´neuc, e e Entropy Giuliano Antoniol Introduction th H Our study Dispersion measures Our study - refined Case study RQ1 – Metric Relevance RQ2 – Relation to Faults ? Conclusions and future work th CC Context Coverage 7 / 16
- 9. Physical and Conceptual Dispersion measures Identifier Dispersion Aggregated metric - numHEHCC Venera (3/3) Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Gu´h´neuc, e e Entropy Giuliano Antoniol Introduction th H Our study Dispersion measures Our study - refined Case study RQ1 – Metric Relevance H: used in few identifiers RQ2 – Relation to Faults CC: used in similar contexts Conclusions and future work th CC Context Coverage 7 / 16
- 10. Physical and Conceptual Dispersion measures Identifier Dispersion Aggregated metric - numHEHCC Venera (3/3) Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Gu´h´neuc, e e Entropy Giuliano Antoniol ? Introduction th H Our study Dispersion measures Our study - refined Case study RQ1 – Metric Relevance RQ2 – Relation to Faults Conclusions and future work th CC Context Coverage 7 / 16
- 11. Physical and Conceptual Dispersion measures Identifier Dispersion Aggregated metric - numHEHCC Venera (3/3) Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Gu´h´neuc, e e Entropy Giuliano Antoniol H: used in many identifiers Introduction CC: used in similar contexts th H Our study Dispersion measures Our study - refined Case study RQ1 – Metric Relevance RQ2 – Relation to Faults Conclusions and future work th CC Context Coverage 7 / 16
- 12. Physical and Conceptual Dispersion measures Identifier Dispersion Aggregated metric - numHEHCC Venera (3/3) Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Gu´h´neuc, e e Entropy Giuliano Antoniol Introduction th H Our study Dispersion measures Our study - refined Case study ? RQ1 – Metric Relevance RQ2 – Relation to Faults Conclusions and future work th CC Context Coverage 7 / 16
- 13. Physical and Conceptual Dispersion measures Identifier Dispersion Aggregated metric - numHEHCC Venera (3/3) Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Gu´h´neuc, e e Entropy Giuliano Antoniol Introduction th H Our study Dispersion measures Our study - refined Case study RQ1 – Metric Relevance RQ2 – Relation to Faults H: used in few identifiers Conclusions and CC: used in different contexts future work th CC Context Coverage 7 / 16
- 14. Physical and Conceptual Dispersion measures Identifier Dispersion Aggregated metric - numHEHCC Venera (3/3) Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Gu´h´neuc, e e Entropy Giuliano Antoniol ? Introduction th H Our study Dispersion measures Our study - refined Case study RQ1 – Metric Relevance RQ2 – Relation to Faults Conclusions and future work th CC Context Coverage 7 / 16
- 15. Physical and Conceptual Dispersion measures Identifier Dispersion Aggregated metric - numHEHCC Venera (3/3) Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Gu´h´neuc, e e Entropy H: used in many identifiers Giuliano Antoniol CC: used in different contexts Introduction th H Our study Dispersion measures Our study - refined Case study RQ1 – Metric Relevance RQ2 – Relation to Faults Conclusions and future work th CC Context Coverage 7 / 16
- 16. Physical and Conceptual Dispersion measures Identifier Dispersion Aggregated metric - numHEHCC Venera (3/3) Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Gu´h´neuc, e e Entropy H: used in many identifiers Giuliano Antoniol CC: used in different contexts Introduction th ! H Our study Dispersion measures Our study - refined Case study RQ1 – Metric Relevance RQ2 – Relation to Faults Conclusions and future work th CC Context Coverage 7 / 16
- 17. Physical and Conceptual Dispersion measures Identifier Dispersion Aggregated metric - numHEHCC Venera (3/3) Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Gu´h´neuc, e e Entropy H: used in many identifiers Giuliano Antoniol CC: used in different contexts Introduction th ! H Our study Dispersion measures Our study - refined Case study RQ1 – Metric Relevance RQ2 – Relation to Faults Conclusions and future work th CC Context Coverage For each entity, numHEHCC counts the number of such terms 7 / 16
- 18. Physical and Conceptual Our study - refined Identifier Dispersion (1/2) Venera Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Gu´h´neuc, e e Giuliano Antoniol Research question 1 Introduction RQ1 – Metric Relevance: Does numHEHCC capture Our study characteristics different from size? Dispersion measures Our believe: Yes it does, although we expect some Our study - refined overlap. Case study RQ1 – Metric Relevance To this end, we verify the following: RQ2 – Relation to Faults 1. To what extend numHEHCC and size vary together. Conclusions and future work 2. Can size explain numHEHCC ? 3. Does numHEHCC bring additional information to size for fault explanation? 8 / 16
- 19. Physical and Conceptual Our study - refined Identifier Dispersion (2/2) Venera Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Gu´h´neuc, e e Giuliano Antoniol Research question 2 Introduction Our study RQ2 – Relation to Faults: Do term entropy and Dispersion context coverage help to explain the presence of faults measures in an entity? Our study - refined Case study Our believe: Yes it does! RQ1 – Metric Relevance RQ2 – Relation to Faults How? Conclusions and 1. Estimate the risk of being faulty when entities contain future work terms with high entropy and high context coverage. 9 / 16
- 20. Physical and Conceptual Objects Identifier Dispersion Venera Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Gu´h´neuc, e e Giuliano Antoniol Objects Introduction ArgoUML v0.16 – a UML modeling CASE tool. Our study Rhino v1.4R3 – a JavaScript/ECMAScript interpreter Dispersion measures and compiler. Our study - refined Case study RQ1 – Metric Relevance Program LOC # Entities # Terms RQ2 – Relation to Faults ArgoUML 97,946 12,423 2517 Conclusions and future work Rhino 18,163 1,624 949 We consider as entities both methods and attributes. 10 / 16
- 21. Physical and Conceptual Case study Identifier Dispersion RQ1 – Metric Relevance (1/3) Venera Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Results for RQ1 – Metric Relevance Gu´h´neuc, e e Giuliano Antoniol To what extend numHEHCC and size vary together? Introduction Our study Dispersion Correlation between numHEHCC and LOC measures Our study - refined ArgoUML: 40% Rhino: 43% Case study numHEHCC RQ1 – Metric Relevance RQ2 – Relation to Faults Conclusions and future work LOC 11 / 16
- 22. Physical and Conceptual Case study Identifier Dispersion RQ1 – Metric Relevance (2/3) Venera Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Results for RQ1 – Metric Relevance Gu´h´neuc, e e Giuliano Antoniol Can size explain numHEHCC ? Introduction Our study Dispersion measures ArgoUML: 17% Our study - refined Rhino: 19% Case study RQ1 – Metric Relevance RQ2 – Relation to Faults Conclusions and future work Composition of numHEHCC. 12 / 16
- 23. Physical and Conceptual Case study Identifier Dispersion RQ1 – Metric Relevance (3/3) Venera Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l Gu´h´neuc, e e e Results for RQ1 – Metric Relevance (cont’d) Giuliano Antoniol Introduction Does numHEHCC bring additional information to size Our study for fault explanation? Dispersion measures Variables Coefficients p-values Intercept -1.688e+00 2e − 16 Our study - refined LOC 7.703e-03 8.34e − 10 Case study MArgoUML RQ1 – Metric Relevance numHEHCC 7.490e-02 1.42e − 05 RQ2 – Relation to Faults LOC:numHEHCC -2.819e-04 0.000211 Conclusions and future work Intercept -4.9625130 2e − 16 LOC 0.0041486 0.17100 MRhino numHEHCC 0.2446853 0.00310 LOC:numHEHCC -0.0004976 0.29788 13 / 16
- 24. Physical and Conceptual Case study Identifier Dispersion Results for RQ2 – Relation to Faults (1/1) Venera Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Gu´h´neuc, e e The risk of being faulty when entities contain terms Giuliano Antoniol with high entropy and high context coverage. Introduction Our study All entities Dispersion measures Our study - refined Case study RQ1 – Metric Relevance RQ2 – Relation to Faults Conclusions and future work 14 / 16
- 25. Physical and Conceptual Case study Identifier Dispersion Results for RQ2 – Relation to Faults (1/1) Venera Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Gu´h´neuc, e e The risk of being faulty when entities contain terms Giuliano Antoniol with high entropy and high context coverage. Introduction Our study All entities Dispersion measures Our study - refined Case study RQ1 – Metric Relevance RQ2 – Relation to Faults Conclusions and future work 14 / 16
- 26. Physical and Conceptual Case study Identifier Dispersion Results for RQ2 – Relation to Faults (1/1) Venera Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Gu´h´neuc, e e The risk of being faulty when entities contain terms Giuliano Antoniol with high entropy and high context coverage. Introduction Our study All entities Dispersion measures Our study - refined 10% of the entities Case study RQ1 – Metric Relevance RQ2 – Relation to Faults numHEHCC Conclusions and future work 14 / 16
- 27. Physical and Conceptual Case study Identifier Dispersion Results for RQ2 – Relation to Faults (1/1) Venera Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Gu´h´neuc, e e The risk of being faulty when entities contain terms Giuliano Antoniol with high entropy and high context coverage. Introduction Our study All entities Dispersion measures Our study - refined 10% of the entities Case study RQ1 – Metric Relevance RQ2 – Relation to Faults numHEHCC Conclusions and future work Risk of being faulty? 14 / 16
- 28. Physical and Conceptual Case study Identifier Dispersion Results for RQ2 – Relation to Faults (1/1) Venera Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Gu´h´neuc, e e The risk of being faulty when entities contain terms Giuliano Antoniol with high entropy and high context coverage. Introduction Our study All entities Dispersion measures Our study - refined 10% of the entities Case study RQ1 – Metric Relevance RQ2 – Relation to Faults numHEHCC Conclusions and future work Risk of being faulty? ArgoUML: 2 x higher Rhino: 6 x higher 14 / 16
- 29. Physical and Conceptual Conclusions and future work Identifier Dispersion Venera Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Conclusions Gu´h´neuc, e e Giuliano Antoniol Entropy and context coverage, together, capture Introduction characteristics different from size! Our study Entropy and context coverage, together, help to explain Dispersion the presence of faults in entities! measures Our study - refined Case study Future directions RQ1 – Metric Relevance RQ2 – Relation to Faults Replicate the study to other systems. Conclusions and Use entropy and context coverage to suggest future work refactoring. Study the impact of lexicon evolution on entropy and context coverage. 15 / 16
- 30. Physical and Conceptual Thank you! Identifier Dispersion Venera Arnaoudova, Laleh Eshkevari, Rocco Oliveto, Yann-Ga¨l e Gu´h´neuc, e e Giuliano Antoniol Introduction Our study Dispersion measures Questions? Our study - refined Case study RQ1 – Metric Relevance RQ2 – Relation to Faults Conclusions and future work 16 / 16
- 31. Physical and Conceptual Binkley, D., Davis, M., Lawrie, D., and Morrell, C. Identifier Dispersion (2009). Venera To CamelCase or Under score. Arnaoudova, Laleh Eshkevari, Rocco In Proceedings of 17th IEEE International Conference on Oliveto, Yann-Ga¨l Gu´h´neuc, e e e Program Comprehension. IEEE CS Press. Giuliano Antoniol Butler, S., Wermelinger, M., Yu, Y., and Sharp, H. Introduction (2009). Our study Relating identifier naming flaws and code quality: An Dispersion measures empirical study. Our study - refined In Proceedings of the 16th Working Conference on Case study Reverse Engineering, pages 31–35. IEEE CS Press. RQ1 – Metric Relevance RQ2 – Relation to Faults Deissenboeck, F. and Pizka, M. (2006). Conclusions and future work Concise and consistent naming. Software Quality Journal, 14(3):261–282. Gyim´thy, T., Ferenc, R., and Siket, I. (2005). o Empirical validation of object-oriented metrics on open source software for fault prediction. 16 / 16
- 32. Physical and IEEE Transactions on Software Engineering, Conceptual Identifier 31(10):897–910. Dispersion Venera Haiduc, S. and Marcus, A. (2008). Arnaoudova, Laleh Eshkevari, Rocco On the use of domain terms in source code. Oliveto, Yann-Ga¨l Gu´h´neuc, e e e In Proceedings of 16th IEEE International Conference on Giuliano Antoniol Program Comprehension, pages 113–122. IEEE CS Introduction Press. Our study Liu, Y., Poshyvanyk, D., Ferenc, R., Gyim´thy, T., and o Dispersion measures Chrisochoides, N. (2009). Our study - refined Modelling class cohesion as mixtures of latent topics. Case study In Proceedings of 25th IEEE International Conference on RQ1 – Metric Relevance RQ2 – Relation to Faults Software Maintenance, pages 233–242, Edmonton, Conclusions and Canada. IEEE CS Press. future work Marcus, A., Poshyvanyk, D., and Ferenc, R. (2008). Using the conceptual cohesion of classes for fault prediction in object-oriented systems. IEEE Transactions on Software Engineering, 34(2):287–300. 16 / 16
- 33. Physical and Conceptual Poshyvanyk, D. and Marcus, A. (2006). Identifier Dispersion The conceptual coupling metrics for object-oriented Venera systems. Arnaoudova, Laleh Eshkevari, Rocco In Proceedings of 22nd IEEE International Conference on Oliveto, Yann-Ga¨l Gu´h´neuc, e e e Software Maintenance, pages 469 – 478. IEEE CS Press. Giuliano Antoniol Takang, A., Grubb, P., and Macredie, R. (1996). Introduction The effects of comments and identifier names on Our study program comprehensibility: an experiential study. Dispersion measures Journal of Program Languages, 4(3):143–167. Our study - refined Zimmermann, T., Premraj, R., and Zeller, A. (2007). Case study RQ1 – Metric Relevance Predicting defects for eclipse. RQ2 – Relation to Faults In Proceedings of the Third International Workshop on Conclusions and future work Predictor Models in Software Engineering. 16 / 16