Observing Activity Patterns in Software Development

Analysing modern, highly complex software projects by observing how their metrics change over time can yield surprising yet useful insights. While most modern software evolution tools provide a good summary on how these metrics develop, they usually do not pay attention to the source contributors at all. This thesis presents a way to connect metric altering commits in Version Control Systems back to their author. By observing every author’s way of coding, a much deeper insight into a software development process can be provided. Bad habits of developers can be discovered and potentially prevented. In turn, this can lead to reduced effort and costs during the development and maintenance of software in the future. A Java implementation to analyse these aspects of software development and maintenance is presented. It extracts Version Control System data including file, module, and function metrics while focusing on the actual committers who are responsible for them. To show the usefulness of this data, a case study on the modern software project Amarok 2 is provided that contains 32232 Git revisions in the master branch alone.