Feature location through the combination of run-time architecture models and information retrieval

Abstract

Eighty percent of the lifetime of a system is spent on maintenance activities. Feature location is one of the most important and common activities performed by developers during software maintenance. This work presents our approach for performing feature location by leveraging the use of architecture models at run-time. Specifically, the execution information is collected in the architecture model at run-time. Then, our approach performs an Information Retrieval technique at the model level. We have evaluated our approach in a Smart Hotel with its architecture model at run-time. We compared our architecture-model-based approach with a source-code-based approach. The rankings produced by the approaches indicate that since models are on a higher abstraction level than source code, they provide more accurate results. Our architecture-model-based approach ranks the relevant elements in the top ten positions of the ranking in 84% of the cases; in the top positions in the ranking of the source-code-based approach, there are false positives associated with some programming patterns and true positives are spread between positions 12 and 100.