Evaluating the influence of scope on feature location

Abstract

[EN] Context: Feature Location (FL) is a widespread technique that is used to maintain and evolve a software product. FL is also helpful in reengineering a family of software products into a Software Product Line (SPL). Despite the popularity of FL, there is no study that evaluates the influence of scope (single product or product family) when engineers perform FL.

Objective: The goal of this paper is to compare the performance, productivity, and perceived difficulty of manual FL when scope changes from a single product to a product family. Method:

We conducted a crossover experiment to compare the performance, productivity, and perceived difficulty of manual FL when scope changes. The experimental objects are extracted from a real-world SPL that uses a Domain-Specific Language to generate the firmware of its products.

Results: Performance and productivity decrease significantly when engineers locate features in a product family regardless of their experience. For these variables the impact of the FL Scope is medium¿large. On contrast, for perceived difficulty, the magnitude of the difference is moderate and is not significant.

Conclusions: While performance and productivity decrease significantly when engineers locate features in a product family, the difficulty they perceive does not predict the significant worsening of the results. Our work also identifies strengths and weaknesses in FL. This can help in developing better FL approaches and test cases for evaluation.