Francisco J. Garcia COMBINED VISUALIZATION OF STRUCTURAL AND ANALYSIS

Transcription

1 Antonio Gonzalez, Roberto Theron, AlexandruTeleaand Francisco J. Garcia COMBINED VISUALIZATION OF STRUCTURAL AND METRIC INFORMATION FOR SOFTWARE EVOLUTION ANALYSIS

2 Combined Visualization of Structural and Metric Information for Software Evolution Analysis Outline Introduction Background Software evolution analysis Aimof thispaper Description of the proposal Discussion of the designs Conclusions Future work Debate questions

3 Introduction Software evolution analysis is concerned with understanding software changes, their causes, and their effects. Understanding software changes require the use of tools for static and dynamic analysis for revealing the structural characteristics of program and its dynamic behavior. understand the function of individual components of the system and their causal relation. predict the code segments affected when changing other parts of the system. the ability for finding errors causes and details for helping determine the feasibility of proposed changes and notify management of any anticipated problems. Cause effect relation toestablishthescopeofachange,topredict potential ripple effects and to trace data and control flow. the measurement of software quality and maintainability evolution is of great importance due to the effects that produce changes during the maintenance of systems.

4 Background Software ConfigurationManagement (SCM) controls the evolution of complex systems taking into consideration Communication at every level of the organization Changes of code and documentation. SCM toolsmustprovide Management of the component database, Concurrency Collaboration, Recording changes Time, date, modules affected, modification duration, who did the change. However, in spiteof therichnessof this data source there is an important lack of mechanisms to convey, how software quality evolves in a particular project, numerous visualizations have been proposed to get insight in the evolution of metrics and structural data, but combining several such attributes in scalable, effective, and efficient views is still an ongoing endeavor.

5 Software evolution analysis SE analysis enable project managers decision making affected by factors such as the dynamics of software quality measured by quality metrics(i.e. complexity, cohesion and coupling). SE analysis support software developers process improvement, fault prediction, productivity estimation, comparing the actual and desired architectures of a product, and planning future development activities. Both groups require comparing and correlating the evolution of structural (i.e. project containment trees, class hierarchies, and entity dependency graphs) and metric information from a software project.

6 Aim of this paper In this paper, we propose and discuss four visualization designs for: the exploration and comparison of software project revisions, project structure and class hierarchy data and the correlation of structural data with metrics defined at revision and class level, as well as associating structure elements and source code.

7 Description Timeline view This visualization design uses a modified circular ring chart layout to show an entire overview of the time dimension of a project. Concentric rings show the different time scales that record change events, from coarse (years, outer ring) to fine (hours or finer, innermost ring). This type of layout compactly presents large quantities of data and provides an overview + detail view.

8 Description Structure evolution view This structure is a variation of the binary H-V tree for the representation of n-ary trees. The advantage of this layout is a very compact usage of the 2D space. A second advantage of this structure visualization is that, combined with interaction, it allows the side-by-side comparison of a moderate number of hierarchies extracted from different project revisions.

9 Description Metrics view Project-wide metrics are useful to monitor overall system quality trends. Class-level metrics help more fine-grained development tasks. Besides metric values the classlevel view a modified representation also indicates the appearance, deletion, or readdition of a class to a revision.

10 Description Indirect class coupling integrating source code viewing This visualization design addresses the representation of indirect coupling integrated with source code viewing. It allows seeing the classes and methods that affect the tasks carried out by a particular class.

11 Discussion of the designs

12 Future work Improvement of the software evolution analysis engine. The proposal of additional visualization designs for providing more details at the source code level. Validation of the proposal by analyzing several software projects by comparison with existing evolution visualizations.

13 Conclusions The four visualization design focus on different tasks and use- cases: overview of the project commits structure and related metrics (timeline view) comparison of evolving package or class hierarchy structures over time (structure evolution view) trend analysis of metrics (metric view) and detailed code inspection (visualization of the indirect class coupling integrating source code viewing).

14 Conclusions The views are linked by interaction, selection, and navigation functions. Although a formal evaluation lacks this type of combination provides added value in cross-role tasks, such as: managers that need to overview, discuss and report activity and project quality correlations or developers that need to learn new code bases, development history, and track structural changes.

15 Questions?

16 Debate question 1 The software evolution analysis process makes available several interesting information details which are useful for software maintainability. Which results of the software evolution analysis process should be of special consideration when designing a visual analytics tool to assist developers and managers of software projects during the maintenance phase of the systems?

17 Debate questions 2 Which software quality and maintainability metrics could be considered essential for supporting developers and managers of software projects during the maintenance phase of the systems?

18 Thanks!