1. Robert Tairas (INRIA & EMN) Ferosh Jacob (University of Alabama) Jeff Gray (University of Alabama) International Workshop on Software Clones (IWSC) – May 23, 2011 Representing Clones in a Localized Manner © AtlanMod | atlanmod-contact@mines-nantes.fr This material is based upon work supported by the National Science Foundation under Grant No. CCF-0702764

2. Problem: Clone Comprehension  Clones in a group can be scattered – Within a large source file or in several files – A programmer must view clones in each location 2 Cloned Code

3. Representation within Source Editor  From the refactoring perspective: – Eclipse refactoring requires multiple dialog boxes – Separation between editor and refactoring tasks – A solution: visualize refactoring changes directly in the source editor 3 Screen shot of Refactor! Pro † † Refactor! Pro, 2010

4. Localized Clone Representation  Represent a clone group in a localized manner directly in the source editor – Parameterized differences visualized in representation 4 if (!delete(file)) { String message = "Unable to delete file " + file.getAbsolutePath(); if (failonerror) { throw new BuildException(message); } else { log(message, quiet ? Project.MSG_VERBOSE : Project.MSG_WARN); } if (!delete(file)) { String message = "Unable to delete file " + file.getAbsolutePath(); if (failonerror) { throw new BuildException(message); } else { log(message, quiet ? Project.MSG_VERBOSE : Project.MSG_WARN); } if (!delete(f)) { String message = "Unable to delete file " + f.getAbsolutePath(); if (failonerror) { throw new BuildException(message); } else { log(message, quiet ? Project.MSG_VERBOSE : Project.MSG_WARN); } if (!delete(f)) { String message = "Unable to delete file " + f.getAbsolutePath(); if (failonerror) { throw new BuildException(message); } else { log(message, quiet ? Project.MSG_VERBOSE : Project.MSG_WARN); } if (!delete(f)) { String message = "Unable to delete file " + f.getAbsolutePath(); if (failonerror) { throw new BuildException(message); } else { log(message, quiet ? Project.MSG_VERBOSE : Project.MSG_WARN); } if (!delete(f)) { String message = "Unable to delete file " + f.getAbsolutePath(); if (failonerror) { throw new BuildException(message); } else { log(message, quiet ? Project.MSG_VERBOSE : Project.MSG_WARN); } if (!delete(dir)) { String message = "Unable to delete directory " + dir.getAbsolutePath(); if (failonerror) { throw new BuildException(message); } else { log(message, quiet ? Project.MSG_VERBOSE : Project.MSG_WARN); } if (!delete(dir)) { String message = "Unable to delete directory " + dir.getAbsolutePath(); if (failonerror) { throw new BuildException(message); } else { log(message, quiet ? Project.MSG_VERBOSE : Project.MSG_WARN); }

5. Displaying Clones in a Localized Manner 5 Original Source Code User Selects a Clone Group Clone Group Generate Suffix Tree Generate Suffix Tree Clone Differences Localized Representation Generate Representation Generate Representation  Determine differences among the clones  Differences based on first-level statement comparisons  Determine differences among the clones  Differences based on first-level statement comparisons  Performed by third-party tool Clone Groups Clone Detection  Localized representation is displayed after a user selects a clone group  Representation “re-calculated” when a different clone group is selected  Localized representation is displayed after a user selects a clone group  Representation “re-calculated” when a different clone group is selected

6. Detecting Parameterized Elements  A suffix tree is generated on the AST nodes representing the statements of a group of clones  Elements in nodes containing allowed differences are mapped together 6 Stmt1 $ $ Clone 1 $ $ # # Stmt1 Stmt2 Stmt1 # # Clone 2 Stmt2 Excerpt of suffix tree file → dir file.getAbsolutePath()dir.getAbsolutePath() Parameterized elements mapped

7. Statement Similarity Levels  Comparing two statements of two clones – Level 1: Corresponding nodes match each other exactly – Level 2: Corresponding nodes are identical, but can contain allowed parameterized differences  i.e., MethodInvocation, NumberLiteral, QualifiedName, SimpleName, or StringLiteral – Level 3: Corresponding nodes are not identical, but both correspond to types from the Level 2 comparison  e.g., a MethodInvocation is matched with a SimpleName. 7

8. Similarity Levels in Representation  Statement 1 in Clone 1 – Exact matching nodes with Statement 1 in Clone 2 8 Clone 2

9. Similarity Levels in Representation  Statement 2 in Clone 1 – Non-matching nodes with Statements 2 and 3 in Clone 2 9 Clone 2

10. Similarity Levels in Representation  Statement 3 in Clone 1 – Parameterized identical and non-identical nodes with Statement 4 in Clone 2 10 Clone 2

11. Similarity Levels in Representation  Statement 4 in Clone 1 – Non-matching nodes with Statement 5 in Clone 2 11 Clone 2

12. Example Representations  Sub-groups of clones – Tighter similarities: Clones 1 and 4 vs. Clones 2 and 3 12 for (int i = 0; i < params.length; i++) { if (CONTAINS_KEY.equals(params[i].getType())) { contains.addElement(params[i].getValue()); } for (int i = 0; i < params.length; i++) { if (CONTAINS_KEY.equals(params[i].getType())) { contains.addElement(params[i].getValue()); } Clone 1 for (int i = 0; i < params.length; i++) { if (COMMENTS_KEY.equals(params[i].getType())) { comments.addElement(params[i].getValue()); } for (int i = 0; i < params.length; i++) { if (COMMENTS_KEY.equals(params[i].getType())) { comments.addElement(params[i].getValue()); } Clone 4 for (int i = 0; i < params.length; i++) { if (PREFIX_KEY.equals(params[i].getName())) { prefix = params[i].getValue(); break; } for (int i = 0; i < params.length; i++) { if (PREFIX_KEY.equals(params[i].getName())) { prefix = params[i].getValue(); break; } Clone 2 for (int i = 0; i < params.length; i++) { if (LINE_BREAKS_KEY.equals(params[i].getName())) { userDefinedLineBreaks = params[i].getValue(); break; } for (int i = 0; i < params.length; i++) { if (LINE_BREAKS_KEY.equals(params[i].getName())) { userDefinedLineBreaks = params[i].getValue(); break; } Clone 3

13. Clone Properties Based on Visualizations 13 Quick summary of clone differences in large clones Quick summary of neighboring clones Identifying an “oddball” clone

14. Evaluation: Fully Representing Clones  Considers the number of clone groups (i.e., #CG) that can be appropriately represented – Evaluated on multiple open source Java projects 14 Project#CGExact (%)Param (%)StmtDiff (%)Mixed (%) Apache Ant 1.6.542961 (14%)152 (35%)131 (31%)85 (20%) ArgoUML 0.2665061 (9%)214 (33%)124 (19%)251 (39%) Jakarta-JMeter 2.3.237777 (20%)158 (42%)71 (19%) JBoss AOP 2.1.515951 (32%)81 (51%)14 (9%)13 (8%) JFreeChart 1.0.10847151 (18%)415 (49%)168 (20%)113 (13%) JRuby 1.4.0318113 (36%)70 (22%)63 (20%)72 (23%) EMF 2.4.128554 (19%)136 (48%)52 (18%)42 (15%) JEdit 4.234591 (26%)120 (35%)88 (26%)46 (13%) Squirrel-SQL 3.0.342878 (18%)164 (38%)70 (16%)116 (27%)

15. Evaluation: Fully Representing Clones  Exact  Clones that match each other exactly 15 Project#CGExact (%)Param (%)StmtDiff (%)Mixed (%) Apache Ant 1.6.542961 (14%)152 (35%)131 (31%)85 (20%) ArgoUML 0.2665061 (9%)214 (33%)124 (19%)251 (39%) Jakarta-JMeter 2.3.237777 (20%)158 (42%)71 (19%) JBoss AOP 2.1.515951 (32%)81 (51%)14 (9%)13 (8%) JFreeChart 1.0.10847151 (18%)415 (49%)168 (20%)113 (13%) JRuby 1.4.0318113 (36%)70 (22%)63 (20%)72 (23%) EMF 2.4.128554 (19%)136 (48%)52 (18%)42 (15%) JEdit 4.234591 (26%)120 (35%)88 (26%)46 (13%) Squirrel-SQL 3.0.342878 (18%)164 (38%)70 (16%)116 (27%)

16. Evaluation: Fully Representing Clones  Param  Groups with parameterized differences – Majority of the cases except ArgoUML and JRuby – Four cases almost half of the instances 16 Project#CGExact (%)Param (%)StmtDiff (%)Mixed (%) Apache Ant 1.6.542961 (14%)152 (35%)131 (31%)85 (20%) ArgoUML 0.2665061 (9%)214 (33%)124 (19%)251 (39%) Jakarta-JMeter 2.3.237777 (20%)158 (42%)71 (19%) JBoss AOP 2.1.515951 (32%)81 (51%)14 (9%)13 (8%) JFreeChart 1.0.10847151 (18%)415 (49%)168 (20%)113 (13%) JRuby 1.4.0318113 (36%)70 (22%)63 (20%)72 (23%) EMF 2.4.128554 (19%)136 (48%)52 (18%)42 (15%) JEdit 4.234591 (26%)120 (35%)88 (26%)46 (13%) Squirrel-SQL 3.0.342878 (18%)164 (38%)70 (16%)116 (27%)

17. Evaluation: Fully Representing Clones  StmtDiff  Groups with statement differences  Mixed  Groups containing both Param and StmtDiff 17 Project#CGExact (%)Param (%)StmtDiff (%)Mixed (%) Apache Ant 1.6.542961 (14%)152 (35%)131 (31%)85 (20%) ArgoUML 0.2665061 (9%)214 (33%)124 (19%)251 (39%) Jakarta-JMeter 2.3.237777 (20%)158 (42%)71 (19%) JBoss AOP 2.1.515951 (32%)81 (51%)14 (9%)13 (8%) JFreeChart 1.0.10847151 (18%)415 (49%)168 (20%)113 (13%) JRuby 1.4.0318113 (36%)70 (22%)63 (20%)72 (23%) EMF 2.4.128554 (19%)136 (48%)52 (18%)42 (15%) JEdit 4.234591 (26%)120 (35%)88 (26%)46 (13%) Squirrel-SQL 3.0.342878 (18%)164 (38%)70 (16%)116 (27%)

18. Summary  Clone group representation localized on just one clone instance directly in the source editor – Provides a quick summary of the clone properties – No need to open every occurrence of each clone – Evaluated artifacts show that > 50% of groups are fully represented  Remaining groups with non-supported parameterized elements still identified appropriately  Future work – Recognize more parameterized elements to reduce the number of non-matched statements – Determine the extent that the representation can be utilized without becoming just cluttered and less useful 18

19. Key Need Emerging from this Work  Many clone detection and analysis tools present results in a proprietary format  This makes it very challenging to chain tools together and to build new functionality  A strong need is a common standard format for representing results of clone tools  Questions: – Why is it that such a standard is not emerging as a widespread need for this community? – What is needed to get tool providers to consider efforts like the Rich Clone Format (RCF)?  http://www.softwareclones.org/rcf.php http://www.softwareclones.org/rcf.php 19

20. Thank You  Personal: – http://www.emn.fr/z-info/atlanmod/index.php/User:Tairas  Code Clones Literature: – http://www.cis.uab.edu/tairasr/clones/literature  Code Clones Literature (RSS Feed): – http://www.cis.uab.edu/tairasr/clones/literature/clones.xml 20