Interactive Exploration of Collaborative Software-Development Data

Abstract

Modern collaborative software-development tools generate a rich data record, over the lifecycle of the project, which can be analyzed to provide team members and managers with insights into the performance and contributions of individual members and the overall team dynamic. This data can be analyzed from different perspectives, sliced and diced across different dimensions, and visualized in different ways. Frequently the most useful analysis depends on the actual data, which makes the design of single authoritative visualization a challenge. In this paper we describe an analysis and visualization tool that supports the flexible run-time mapping of such a data record to a number of alternative visualizations. We have used our framework to analyze and gain an understanding of how individuals work within their teams and how teams differ in their work on these term projects.URL (screencast and system): http://hypatia.cs.ualberta.ca/ collab/Visualizations/Dashboard/main/splash/splash.html I. THE PROBLEM CONTEXT Collaborative software-engineering projects rely on the use of tools for the coordination of the work among the team members and the management of their work products. Systems such as CVS, Subversion and GitHub enable team members to work in an asynchronous yet coordinated manner, to share their work products, and even to backtrack over erroneous commits to their codebase. Issue tracking systems enable developers to inform each other about their activities, and to prioritize and track progress on pending milestones and outstanding bugs. Mailing lists and wikis offer more flexible channels through which to communicate about schedules, pending work and documentation, and to broadcast information to all team members. Instant messengers and IRC chat channels provide a real-time communication when developers are working synchronously but not in the same physical location.The artifacts produced by these tools provide a rich history of the project as well as clues to the communication and development patterns of the individual developers and the team as whole. Yet understanding this collection of data is not straightforward and can be time consuming. A number of tools have been developed to analyze and visualize this data, yet more work is necessary especially when one considers the question of "understanding individual contribution and team dynamics". This question is interesting for all software projects but especially relevant for teams of novice developers with immature work practices and varied development knowledge and skills, as is the case with academic project-based courses.Motivated by the project-course scenario, we have developed a visual tool, the "Collaboratorium Dashboard", to help instructors navigate through, and interact with, their student teams' data through the web browser.Aware of a number of special-purpose source-code repository visualizations, and having developed some ourselves, in this project we set out to define a general framework, for flexibly creating visualizations of any dataset. The...