Using Provenance to Trace Software Development Processes

Kurzfassung

Today's software development processes are complex in their nature. Simple development process models, such as the waterfall model, have been replaced by iterative methods, such as the Rational Unied Process (RUP), or more exible agile approaches. Furthermore many tools are used in these processes, such as version control systems, issue trackers or continuous integration frameworks. In all phases of the software development, from planning to the implementation and nal delivery, numerous interactions occur between developers, the tools they use while developing and automatically between dierent tools. Examples of these interactions include: i) discussion about a feature request, ii) entering or changing requirements in an issue tracking system iii) automatic code style checks during a check-in. Information about these processes is, if available, distributed over the dierent tools being used. Version control systems feature a history of all les and their editors, issue tracking systems a list of all comments for an issue. Still, the missing link between these dierent tools makes it either impossible to draw conclusions from these data, or very time-consuming, considering the immense amount of available data. Many questions appearing on a daily basis cannot be answered, e.g., "To which requirement does this commit belong?" or "Who is responsible for a failing unit-test?". This master thesis aims at making it possible to follow and reproduce the entire process of developing a software product, being able to easily answer the above questions at any point in the development process. An example process is selected and transformed into a specic model, holding all required data. As this data should not be inserted manually into the model by developers, but recorded automatically, the model must be integrated into a distributed software engineering tool suite (version control system, bug tracking, integrated development environment, etc.). Furthermore, it must be possible to perform automatic audits and analysis on the collected data. The foundations of this thesis have been established in the research area of provenance. Provenance is dened as the origin or history of data. General models to describe and store data, query the data for information and methods for adapting processes to the model have been developed and are used as a basis.