Paper: Measuring Disruption from Software Evolution Activities Using Graph-Based Metrics Authors: Prashant Paymal, Rajvardhan Patil, Sanjukta Bhowmick and Harvey Siy Session: Early Research Achievements Track Session 3: Managing and Supporting Software Maintenance Activities

1. Measuring Disruption from Software
Evolution Activities Using Graph-
Based Metrics
Prashant Paymal, Rajvardhan Patil, Sanjukta Bhowmick, Harvey Siy
Department of Computer Science,
University of Nebraska at Omaha

2. Introduction
• Real world software systems have large numbers
of components (e.g. classes, functions, etc.)
• It is difficult to get a quick summary of how
system evolved after a major change such as
perfective maintenance activity or new software
release

3. Case Study
Version Date Commit Messages
V1 3/9/2001 Merge to JHotDraw 5.2 (using JFC/Swing GUI components)
V2 10/24/2001 Before merge for version 5.3 (dnd, undo…) merge dnd
(before 5.3)
V3 8/4/2002 After various merges… (before 5.4 relaease)
V4 11/8/2002 Refactor to use StandardStorageFormat as a superclass
V5 5/8/2003 Refactoring of Cursor. – java.awt.Cursor(class) has been
systematically replaced
V6 1/9/2004 After renaming the CH.ifa.draw to org.jhotdraw
• Our case study consists of six versions of JHotDraw from
March 2001 to January 2004

4. Network Construction
• Extracted relationships from these versions
(inheritance, implementation, method calls and class member
access, object declaration and instantiation)
• Network was created by connecting class
dependencies, where each edge (u, v) is a
dependency from class „u‟ to class „v‟

5. Vertex Properties
• Degree Distribution
▫ Frequency of vertices per degree, scale
free for most real world networks
• Clustering Coefficient
▫ Connections between neighbors
• Betweenness Centrality
▫ Ratio of shortest paths through a
vertex
• Articulation Points
▫ It‟s removal would cause the network to
become disconnected

6. • Network representing Version 1,
▫ Lighter Nodes: High Betweenness Centrality
▫ Larger Nodes: High Clustering Coefficient

7. Objective
• Extract key combinatorial properties from these
six networks that would enable us to detect
evolutionary characteristics such as
▫ Points of significant change in the software
▫ How these changes affect crucial classes in the
network

8. Change in Vertex Properties
• All properties increased with version number

9. Correlation Between Properties
▫ Positive correlation between degree and betweenness centrality
▫ Correlation between clustering coefficient and betweenness
centrality changes across versions

10. Disruption in Values and Rank
• We examine how the relationships between
these properties changed from one version to the
next

11. Disruption in Values and Rank

12. Disruption in Values and Rank

13. Identifying Crucial Vertices
• High
▫ If vertex has high rank (within top 25) in at least one of the
following categories
• Extra High
▫ If vertex has high rank in at least two categories
• Low
▫ If vertex has zero value for any one vertex based properties
and is not marked as a High vertex
• Extra Low
▫ If it has zero value for both betweenness centrality and
clustering coefficient
• (High Betweenness Centrality, High Indegree, High Outdegree,
High Clustering Coefficient / Articulation Point)

14. Percentage Breakdown of All Vertices
in Each Version

15. Percentage Breakdown of Vertices
(Common to All Versions)
Other
Extra
Low
Low
Extra
High
High

16. Analysis of Newly Added Vertices

17. Bug Frequencies
• Changes that have the keywords “bug fix” in the change log
• The periods with high percentage are also the periods after the high
disruption

18. Conclusion
• The significant evolutionary changes occur between
Version 2 – Version 3 and Version 4 – Version 5
• The network has grown cumulatively. Newer vertices
tend to get added to the peripheries of the network
• The top 25 ranking of vertices was generally stable
across versions. Important nodes stay important. This
indicates stability in the design.
• The bug frequency is higher after Version 3 and Version
5. The degree of disruption can help explain why bug
incidence increases (future work)

19. Acknowledgement
• Nebraska EPSCoR
• College of IS&T, University of Nebraska at
Omaha

20. Thank you!