Tracing requirements and source code during software development

Traceability supports the software development process in various ways, amongst others, change management, software maintenance and prevention of misunderstandings. Traceability links between requirements and code are vital to support these development activities, e.g., navigating from a requirement to its realization in the code, and vice versa. However, in practice, traceability links between requirements and code are often not created during development because this would require increased development effort. This reduces the possibilities for developers to use these links during development. To address this weakness, this thesis presents an approach that (semi-) automatically captures traceability links between requirements and code during development. We do this by using work items from project management that are typically stored in issue trackers. The presented approach consists of three parts. The first part comprises a TIM consisting of artifacts from three different areas, namely requirements engineering, project management, and code. The TIM also includes the traceability links between them. The second part presents three processes for capturing traceability links between requirements, work items, and code during development. The third part defines an algorithm that automatically infers traceability links between requirements and code based on the interlinked work items. The traceability approach is implemented as an extension to the model-based CASE tool UNICASE, which is called UNICASE Trace Client. Practitioners and researchers have discussed the practice of using work items to capture links between requirements and code, but there has been no systematic study of this practice. This thesis provides a first empirical study based on the application of the presented approach. The approach and its tool support are applied in three different software development projects conducted with undergraduate students. The feasibility and practicability of the presented approach and its tool support are evaluated. The feasibility results indicate that the approach creates correct traceability links between all artifacts with high precision and recall during development. At the same time the practicability results indicate that the subjects found the approach and its tool support easy to use. In a second empirical study we compare the presented approach with an existing technique for the automatic creation of traceability links between requirements and code. The results indicate the presented approach outperforms the existing technique in terms of the quality of the created traceability links.