Metrik-basierte Auswertung von Software-Entwicklungsarchiven zur Prozessbewertung

The development and maintenance of complex software products requires transparency of processes and costs. Measurements during the development process create transparency and enable to assess and continuously monitor development processes. A systematic measurement approach requires an organization-specific infrastructure for the collection and evaluation of measurement results. The necessary effort imposes a barrier especially for small and medium-sized enterprises. A lot of information on development processes is routinely collected in software repositories, for example issue tracking systems or configuration management systems. Evaluation of this data is an alternative to the manual collection of status information. In industrial practice, this data is not sufficiently used for process assessment. The reasons for this are on the one hand lack of methodical support, on the other hand lack of flexible tool support. The goal of the methods and tools developed in this thesis is to make better use of the data collected in software repositories for process assessment. The core of the presented solution is the declarative language ITMS for the specification of metrics on issue tracking systems. This language facilitates a compact and precise description of metrics on a high abstraction level. The presented reference implementation of the language can be flexibly adapted to different software repositories. Moreover the metrics specified in ITMS are easily maintainable, such that an iterative procedure for development and validation of metrics can be applied. To ease the systematic interpretation of measurement results, a meta-model for quality models will be presented. Such a quality model represents the relation between subjective quality characteristics and the measurements. These concepts are demonstrated in a quality model editor and evaluation tool. The evaluation tool supports the classification of measurement results based on empirical comparison data. This facilitates a pragmatic and realistic interpretation of the measurement results. Applicability and scalability of the developed methods and tools are demonstrated with case studies in the context of industrial software development as well as in the context of open source software