InMap: Automated Interactive Code-to-Architecture Mapping Recommendations

Abstract

Reflexion Modelling is a popular method used in industry for Software Architectural Consistency Checking (SACC). However, it involves a mapping step that is manual and tedious. There exist techniques and tools that attempt to automate mapping, yet they are either limited in their approach or they require an initial set of manually pre-mapped entities. This study proposes a novel technique, InMap, that improves the mapping process in reflexion modelling by both providing versatility and eliminating the constraint of needing a set of manually pre-mapped entities in order to automate mapping. Using a software’s architecture descriptions, InMap applies information retrieval concepts to the software’s source code to interactively provide mapping recommendations to an architect. For the six systems InMap was evaluated on, the recommendations it provided achieved an average recall of 0.97, and an average precision of 0.82. InMap also achieved higher, f1-scores in comparison to existing techniques that require premapping. This provides a basis for improving industry tools that use reflexion modelling or similar SACC methods.