The document discusses algorithms for recommending refactoring operations in large software systems. It describes two main algorithms - a clustering-based algorithm and a graph-based algorithm. The clustering-based algorithm identifies groups of similar methods and entities that could be extracted into separate classes. The graph-based algorithm builds a matrix of relationships between methods and identifies strongly related chains of methods that could form candidate classes. Both algorithms were evaluated on open source systems and shown to effectively recommend extract class refactoring opportunities.

1. Recommending Refactoring Operations in Large
Software SystemsPósGraduaçãoemCiênciadaComputaçãoAluno:CarlosEduardoDantasOrientador:MarceloMaia

2. Recommending Refactoring Operations in Large Software SystemsIntroduction
•During its lifecycle, the internal structure of a Software system undergoes continuous modifications;
•Source code quality often decreases;
•Low quality is generally associated with lower productivity, more rework and more effort for developers.

3. Recommending Refactoring Operations in Large Software SystemsRefactoring
•Process of changing Software System to improve its internal structure;
•Help in removing codebad smells as well as antipatterns;
•Improve Architeture and provide better Software extensibility.

4. Recommending Refactoring Operations in Large Software SystemsChallenges on Refactoring code
•Design flaws are not always obvious;
•Its not easy to apply the correct refactoring operation to solve design problems;
•Apply the refactoring solution without changing the external behaviour of the system;
•Manual refactoring is an error-prone task.

5. Recommending Refactoring Operations in Large Software SystemsChallenges on Refactoring code
•Refactoring recommendation systems supports:
Identifying refactoring opportunities;
Designing and applying a refactoring solution.

6. Recommending Refactoring Operations in Large Software SystemsRefactoring operations catalog
•Some refactoring operations, classified on the benefits provided to the source code.
Improving Code Decomposition
Extract Class
Extract Package
Extract Method
Improving names and location of code
Rename method (field)
Move method
Move class
Improving Conformance with OOP principles
Push down field/method
Pull up field/method
Extract/collapse hierarchy

7. Recommending Refactoring Operations in Large Software SystemsAlgorithms exploited to identify refactoring recommendations
RefactoringOperation
Approach
Extract Class
Clustering-basedor Graph-based
Move Method
Heuristic-based or Search-based
Extract Method
Slicing-based
Extract Package
Graph-based
Move Class
Search-based or Heuristic-based
Combination of multiple operations
Search-based
•This presentation will discuss only Extract Class Refactoring approaches.

8. Recommending Refactoring Operations in Large Software SystemsExtract Class Refactoring
•In OOP design, classes are called “God Class”, “System Class” or “Blob Class” when become very large, less cohesive and exhibit high levels of coupling;
•Class should implement only one concept, having only one reason to change;
•Extract class split the responsabilities implemented on Blob classes into different classes with higher cohesion.

9. Recommending Refactoring Operations in Large Software SystemsExtract Class Refactoring with Clustering-Based algorithm
•Clustering methods can identify conceptually meaningful groups of similar entities;
•Clusters may represent cohesive groups of class members;
•Clusters compute an entity set for each attribute and invoke method.

10. Recommending Refactoring Operations in Large Software SystemsClustering-Based algorithm – extract entities

11. Recommending Refactoring Operations in Large Software SystemsClustering-Based algorithm – entity set
a1
name
changeJob
modifyName
getTelephoneNumber
a2
job
changeJob
modifyName
getTelephoneNumber
a3
officeAreaCode
getTelephoneNumber
a4
officeNumber
getTelephoneNumber
m1
changeJob
job
name
m2
modifyName
job
name
m3
getTelephoneNumber
job
name
officeAreaCode
officeAreaNumber

12. Recommending Refactoring Operations in Large Software SystemsClustering-Based algorithm – Jaccard method
•Jaccard method calculate the distance between entities, generating distance matrix.

13. Recommending Refactoring Operations in Large Software SystemsClustering-Based algorithm – cluster deondrogram
•Cluster´s hierarchy is usually represented by a deondrogram.

14. Recommending Refactoring Operations in Large Software SystemsClustering-Based algorithm – Evaluation
Software
Total Suggestions
Assigned Names
Applied
eRisk
37
28
16
SelfPlanner
14
12
9

15. Recommending Refactoring Operations in Large Software SystemsGraph-Based algorithm – class extraction process
•The candidate class is parsed to build a method-by-method matrix;
•NxN matrix where n is the number of methods of class to be refatored;
•Each entry represents the likelihood that method mi and method mj should be in the same class;
•Identify chains of strongly related or coupled methods.

16. Recommending Refactoring Operations in Large Software SystemsGraph-Based algorithm – class extraction process

17. Recommending Refactoring Operations in Large Software SystemsGraph-Based algorithm Method-by-method matrix
•Structural Similarity between Methods;
•Call-based dependence between methods;
•Conceptual similarity between methods.

18. Recommending Refactoring Operations in Large Software SystemsGraph-Based algorithmMethod-by-method matrix

19. Recommending Refactoring Operations in Large Software SystemsGraph-Based algorithmIdentifying candidate chains
Constant Threshold
Variable Threshold

20. Recommending Refactoring Operations in Large Software SystemsGraph-Basedalgorithm – Evaluation
•Created an artificial scenario, where classes with high cohesion of five open source systems have been merged to build classes with low cohesion and many responsibilities

21. Recommending Refactoring Operations in Large Software SystemsGraph-Basedalgorithm – Evaluation
•Constant threshold (0.1,0.2,0.3,0.4)

22. Recommending Refactoring Operations in Large Software SystemsGraph-Basedalgorithm – Evaluation
•Variable threshold (Q1,Q2,Q3)

23. Recommending Refactoring Operations in Large Software SystemsGraph-Basedalgorithm – Evaluation
•Interaction between weight and variable threshold on JHotDraw merging 2 classes

24. Recommending Refactoring Operations in Large Software SystemsGraph-Basedalgorithm – Evaluation
•Interaction between weight and variable threshold on JHotDraw merging 3 classes

25. Recommending Refactoring Operations in Large Software SystemsGraph-Basedalgorithm – Evaluation
•Interaction between weight and system merging 2 classes

26. Recommending Refactoring Operations in Large Software SystemsGraph-Basedalgorithm – Evaluation
•Interaction between weight and system merging 3 classes

27. Recommending Refactoring Operations in Large Software SystemsReferences
•[1] Robillard, Martin P.:, Maalej, Wallid.: Recommendation Systems in Software Engineering. Springer, 2014;
•[2] Fokaefs, M., Tsantalis, N., Chatzigeorgiou, A., Sander, J.: Decomposing object-oriented class modules using an agglomerative clustering technique. In: Proceedings of the IEEE International Conference on Software Maintenance, (2009);
•[3] Fowler, M.: Refactoring: improving the design of existing code. Addison-Wesley, Reading, 1999;

28. Recommending Refactoring Operations in Large Software SystemsReferences
•[4] Fokaefs, M., Tsantalis, N., Stroulia, E., Chatzigeorgiou, A.: Identification and application of extract class refactoringsin object-oriented systems. J. Syst. Software 85(10), 2012;
•[5] Bavota, G., De Lucia, A., Marcus, A., Oliveto, R.: Automating extract class refactoring: a Novel Approach and its Evaluation, 2011;
•[6] Bavota, G., De Lucia, A., Marcus, A., Oliveto, R.: In Medio Stat Virtus: Extract Class Refactoring through Nash Equilibria, 2011.

29. Recommending Refactoring Operations in Large Software SystemsAppendix –Another Graph- Based algorithm example
Method
Instance variables
changeJob
{job,name}
modifyName
{job,name}
getTelephoneNumber
{job,name,officeAreaCode,officeNumber}
CJ
MN
GT
CJ
1
1
0,5
MN
1
1
0,5
GT
0,5
0,5
1
Method
Callmethods
changeJob
{}
modifyName
{}
getTelephoneNumber
{}
CJ
MN
GT
CJ
1
1
0
MN
0
1
0
GT
0
0
1
Method
Conceptual Similarity
changeJob
{}
modifyName
{}
getTelephoneNumber
{}
CJ
MN
GT
CJ
1
0.7
0.1
MN
0.7
1
0.1
GT
0.1
0.1
1

30. Recommending Refactoring Operations in Large Software SystemsAppendix –Another Graph- Based algorithm example
CJ
MN
GT
CJ
1
0.7
0.2
MN
0.7
1
0.2
GT
0.2
0.2
1