SNIAFL: Towards a Static Non-Interactive Approach to Feature Location

To facilitate software maintenance and evolution, a helpfulstep is to locate features concerned in a particular maintenancetask. In the literature, both dynamic and interactive approacheshave been proposed for feature location. In this paper, wepresent a static and non-interactive method for achieving thisobjective. The main idea of our approach is to use theinformation retrieval (IR) technology to reveal the basicconnections between features and computational units in sourcecode. Due to the characteristics of the retrieved connections, weuse a static representation of the source code named BRCG tofurther recover both the relevant and the specific computationalunits for each feature. Furthermore, we recover therelationships among the relevant units for each feature. Apremise of our approach is that programmers should usemeaningful names as identifiers. We perform an experimentalstudy based on a GNU system to evaluate our approach. In theexperimental study, we present the detailed quantitativeexperimental data and give the qualitative analytical results.