CSMR10c.ppt
1 like248 views
This document summarizes a technique for automatically segmenting source code identifiers into meaningful words. It presents a search-based approach inspired by how developers compose identifiers using terms and applying word transformations. The approach uses a dictionary of terms, calculates the distance between an identifier and dictionary words using Dynamic Time Warping, and applies word transformation rules. An evaluation on two systems found it outperformed a simple CamelCase splitter, correctly splitting over 90% of identifiers. Future work is planned to expand the evaluation and enhance heuristics for term selection and transformations.
![Problem Statement
Our Approach in Essence
Developers compose identifiers:
Using terms and words reflecting domain concepts,
developer’s experience, knowledge.
Developers generate contraction via a finite set of
transformation rules:
Drop all vowels, drop prefix, drop suffix, etc.
Mimics developer’s identifiers generation process:
Dictionaries capturing terms and words;
A search-based technique to split exactly any unknown
string;
A distance using Dynamic Time Warping (DTW) for
CSMR 2010, Madrid
continuous speech recognition [H. Ney, 1984].
7/24](https://image.slidesharecdn.com/csmr10c-ppt-130410095153-phpapp01/85/CSMR10c-ppt-7-320.jpg)
.
21/24](https://image.slidesharecdn.com/csmr10c-ppt-130410095153-phpapp01/85/CSMR10c-ppt-21-320.jpg)
![References
[ELK, 2009] E. Enslen, E. Hill, L. Pollock, and K. Vijay-Shanker,
“Mining source code to automatically split identifiers for software
analysis,” Mining Software Repositories, International Workshop on,
vol. 0, pp. 71 - 80, 2009.
[H. Ney, 1984] H. Ney, “The use of a one-stage dynamic programming
algorithm for connected word recognition,” Acoustics, Speech and
Signal Processing, IEEE Transactions on, vol. 32, no. 2, pp. 263 - 271,
Apr 1984.
D. Lawrie, C. Morrell, H. Feild, and D. Binkley, “Effective identifier
names for comprehension and memory,” Innovations in Systems and
Software Engineering, vol. 3, no. 4, pp. 303 - 318, 2007.
D. Lawrie, C. Morrel, H. Feild, and D. Binkley, “What’s in a name? a
study of identifiers,” in Proc. of the International Conference on
Program Comprehension (ICPC), 2006, pp. 3 - 12.
CSMR 2010, Madrid
23/24](https://image.slidesharecdn.com/csmr10c-ppt-130410095153-phpapp01/85/CSMR10c-ppt-23-320.jpg)
CSMR10c.ppt
- 1. Recognizing Words from Source Code Identifiers using Speech Recognition Techniques CSMR 2010, Madrid Nioosha Madani, Latifa Guerrouj, Massimiliano Di Penta, Yann-Gaël Guéhéneuc, and Giuliano Antoniol
- 2. Content Problem Statement Aligning Strings and Words Meta- Meta-heuristic Inspired Approach Technologies Case Study – Research Questions Case Study – Results CSMR 2010, Madrid Conclusion and Future Work 2/24
- 3. Problem Statement The Challenge A few years after deployment, documentation may no longer exist. If it exists, it will be almost surely outdated. My customers desire to change the system, add new functionalities or fix a defect. The only available source of information is the code: Identifiers; Comments. CSMR 2010, Madrid 3/24
- 4. Problem Statement Identifiers Semantic Researchers agree that the identifier semantics are important: Help program comprehension; Suggest clues. Composed identifiers: Camel Case: MyLocalAccount , User_Address Contraction based: pntrctr , usrAdrss , imagEdge Good and possibly known to the developers: hmmm, ixoth , pqrstuvwxyz CSMR 2010, Madrid 4/24
- 5. Problem Statement Words, Terms, Soft, and Hard Words Term: any substring in a compound identifier. Word: an entry in a dictionary (e.g., the English dictionary). Hard words: terms composing an identifier reflecting domain concepts, clearly demarked: baseAddress, baseAddress, user_file Soft words: terms different from the identifier and not clearly demarked (e.g., abbreviation, contraction, etc.): CSMR 2010, Madrid userarea, ptrcntr, userarea, ptrcntr, userGid 5/24
- 6. Problem Statement Current Practices Camel Case-based approaches plus greedy Case- algorithms, e.g., Lawrie et al. 2006, 2007. Samurai by Enslen et al, 2009: Lexicon plus a greedy algorithm; If a contraction is used somewhere in the code then it is likely used in the same context than the original term; Frequency tables of contractions and terms to split composed identifiers. Limitations : Abbreviations not treated, no quantification of how close the match is to the CSMR 2010, Madrid unknown string. 6/24
- 7. Problem Statement Our Approach in Essence Developers compose identifiers: Using terms and words reflecting domain concepts, developer’s experience, knowledge. Developers generate contraction via a finite set of transformation rules: Drop all vowels, drop prefix, drop suffix, etc. Mimics developer’s identifiers generation process: Dictionaries capturing terms and words; A search-based technique to split exactly any unknown string; A distance using Dynamic Time Warping (DTW) for CSMR 2010, Madrid continuous speech recognition [H. Ney, 1984]. 7/24
- 8. Aligning Strings and Words Modified H. Ney DTW 3 5 4 0 U s r 4 5 4 3 2 1 2 3 4 3 4 3 3 1 0 1 3 Dictionary of 3 words r 1 2 2 3 2 2 0 1 2 3 4 5 4 2 1 0 3 4 5 t 2 3 4 3 1 0 1 2 3 4 1 2 3 2 0 1 2 1 2 3 C 3 0 r 3 2 3 3 2 4 5 4 2 2 0 1 2 2 3 3 4 4 P n t 1 0 1 2 2 3 3 2 3 3 0 1 2 3 1 2 2 1 2 2 CSMR 2010, Madrid p n t r c t r u s r Identifier to split : pntrctrusr 8/24
- 9. Meta-heuristic Inspired Approach Word Transformation Rules Constraint: String must remain longer or equal to 3 chars Drop all vowels pointer → pntr Drop a random vowel user → usr Drop a random character pntr → ptr Drop suffix (ing, tion, ed, available → avail ment, able) Drop the last m characters rectangle → rect CSMR 2010, Madrid 9/24
- 10. - Meta-heuristic Inspired Approach -Technologies Overall Splitting (Hill Climbing) Procedure Identifier DTW Match Best Matching Success! Zero Dist? No Select randomly a word with a minimal distance <> 0 Apply a random transformation to the Add transf word to chosen word temporary dictionary Current dictionary yes Discard word Best Matching DTW red Dist ? CSMR 2010, Madrid from temporary Match No dictionary If other transf to apply 10/24
- 11. Case Study – Research Questions Case Study - Research Questions RQ1: What is the percentage of identifiers correctly split by the proposed approach? RQ2: How does the proposed approach perform compared with the Camel Case splitter? RQ3: What percentage of identifiers containing word abbreviations is the approach able to CSMR 2010, Madrid map to dictionary words? 11/24
- 12. Case Study – Results Case Study - Results JHotDraw – Java 16 KLOC 155 files 2,348 identifiers (longer than 2 chars) 957 manually segmented identifiers Lynx – C 174 KLOC 247 files 12,194 identifiers (longer than 2 chars) 3,085 manually segmented identifiers CSMR 2010, Madrid 12/24
- 13. Case Study – Results RQ1 - Percentage of Correct Classifications Splits Ids Single Multiple Errors Systems iteration iterations JHotDraw 957 891 (93%) 920 (95%) 37 Lynx 3,085 2,169 (70%) 2,901 (94%) 271 Typical cases where the approach failed: afaik, ihmo, foobar, fsize … CSMR 2010, Madrid 13/24
- 14. Case Study – Results RQ2 - Camel Case Split Splits Ids Correct Split Errors Systems JHotDraw 957 874 (91%) 83 Lynx 3,085 561 (18%) 2,524 Statistical comparison (Fisher’s exact test) with our approach: Null Hypothesis (H0) : The propotions of correct splittings obtained by the approaches are not significantly <>. • JHotDraw: Odds Ratio = 1.3, p-value = 0.1 CSMR 2010, Madrid • Lynx: Odds Ratio = 60, p-value < 0.001 14/24
- 15. Case Study – Results RQ3 - Percentage of Correctly Split Id (s) Splits Ids Correct Split Errors Systems JHotDraw 957 920 (95%) 37 Lynx 3,085 2,901 (94%) 271 The novel identifier splitting approach perfoms better than the Camel Case splitter. CSMR 2010, Madrid 15/24
- 16. Case Study – Results Multiple Possible Splits - Successes borddec bord decimal bord decision anchorlen anchor length anchor lender drawrect draw rectangle drawroundrect draw round rectangle fillrect fill rectangle javadrawapp java draw apply java draw append netapp net apply net append newlen new length new lender nothingapp nothing apply nothing application addcolumninfo add column information add column inform addlbl add label casecomp case compare case complete Max of 10000 iterations CSMR 2010, Madrid 16/24
- 17. Case Study – Results Multiple Possible Splits - Failures serialversionuid serial version did selectionzordered selection ordered removefrfigurerequestremove remove figure request remove jhotdraw hot draw getvadjustable get bad just able fimagewidth him age width fimageheight him age height writeref write red Max of 10000 iterations DTW does not account for context, syntax or semantic CSMR 2010, Madrid 17/24
- 18. Case Study – Results Discussion - Challenges How can we expand fwrite or pdraw? pdraw? How can we avoid expanding FileLen into File Lender rather than File Length? Length? How can we recognize that ImagEdit has a correct split at distance 1 and not 0? How can we expand/split pqrstuvwxyz? pqrstuvwxyz? CSMR 2010, Madrid 18/24
- 19. Case Study – Results Threats to Validity External validity: We analyzed only two systems; However: different domains, different programming languages. Construct validity: errors may be present in the oracle! We detected 1% error in the first oracle release; We did the best to guess programmer intention but we cannot exclude errors. Reliability validity: replication package available. Internal validity: subjectivity and bias in building the oracle: The same researcher built both oracles; Oracles were validated by other two researchers; Size of oracle large enough to avoid a few percent errors change CSMR 2010, Madrid conclusions. 19/24
- 20. Conclusion and Future Work Conclusion We presented a search-based approach to search- automatically segment source code identifiers. The novel approach is inspired by the developer behavior when composing identifiers. The approach uses a dictionary, a distance computed via DTW, and a set of word transformations. Results on JHotDraw and Lynx show the superiority of the approach over a simple Camel Case splitter. CSMR 2010, Madrid 20/24
- 21. Conclusion and Future Work Future Work We plan to: to: Expand the evaluation to other systems. Introduce enhanced heuristics for term selection and word transformations. Contextualize our search by coupling our algorithm with the approach of Enslen et al. [ELK, 2009](restrict the search to the words used 2009](restrict CSMR 2010, Madrid in the same method, class, or package). 21/24
- 22. Finally… Questions Thank you for your attention CSMR 2010, Madrid 22/24
- 23. References [ELK, 2009] E. Enslen, E. Hill, L. Pollock, and K. Vijay-Shanker, “Mining source code to automatically split identifiers for software analysis,” Mining Software Repositories, International Workshop on, vol. 0, pp. 71 - 80, 2009. [H. Ney, 1984] H. Ney, “The use of a one-stage dynamic programming algorithm for connected word recognition,” Acoustics, Speech and Signal Processing, IEEE Transactions on, vol. 32, no. 2, pp. 263 - 271, Apr 1984. D. Lawrie, C. Morrell, H. Feild, and D. Binkley, “Effective identifier names for comprehension and memory,” Innovations in Systems and Software Engineering, vol. 3, no. 4, pp. 303 - 318, 2007. D. Lawrie, C. Morrel, H. Feild, and D. Binkley, “What’s in a name? a study of identifiers,” in Proc. of the International Conference on Program Comprehension (ICPC), 2006, pp. 3 - 12. CSMR 2010, Madrid 23/24
- 24. Overall Splitting (Hill Climbing) Procedure Best Matching Success! Zero Dist? Identifier DTW Match No Ranked Word List No Yes Improved? Discard word and create new dictionary Temporary Dictionary Dictionary Save word and create new dictionary CSMR 2010, Madrid 24/24