Description
This thesis lifts Git, the most common version-control-system, to the world of massively AI-generated source code, which requires lifting the level of abstraction at which developers work. The thesis project builds on the Git with Features concept for feature-aware version control and extends it toward full product derivation support. The goal is to enable developers to check out product-specific projections—materialized code variants generated from a feature selection—and later commit back their changes in a feature-consistent manner. The work will (1) design a projection model that maps feature selections to concrete code artifacts based on feature–commit associations, (2) implement Git-native commands for creating, editing, and synchronizing derived products, and (3) provide mechanisms for propagating changes back into the platform without losing feature traceability. The resulting approach will be validated through a prototype and empirical evaluation, demonstrating how feature-based product derivation in Git can reduce merge effort, prevent inconsistencies across variants, and support Software Product Line workflows directly in mainstream developer tooling.
Another closely related topic are variation-control systems. Potential concrete topics could be:
- intelligent code merge tool which, for instance, can handle code-alignment issues
- identify side effects of source code using static analysis
- conceive and implement a feature dashboard, showing developers what features exist, where they reside in the code, and show various metrics about features (e.g., scattering degree, tangling degree, lines of feature code)
- create a diff/merge tool for software variants
- a lightweight realization of VTS (variation tracking system)
Contact:
Extent: B.Sc./ M.Sc.