Binary-Search Based Verification of Feature Models

Abstract

The purpose of feature models’ verification is to detect deficiencies in feature models, so as to avoid the transmission of these deficiencies into subsequent core-asset and product development activities. Although many researchers have observed that the verification problem of feature models can be transformed into SAT problems and proposed to resolve this problem based on third-party’s SAT-solver or model-checker tools, few of them point out how to use these third-party tools efficiently. In this paper, we present a binary-search based approach to feature models’ verification. Our motivation is to decrease the number of times a SAT-solver is invoked during the verification of a feature model, and thus improve the verification efficiency. The basic idea is to change feature models’ verification from the linear-search based approach to a binary-search approach, and thereby decrease the number of times to invoke a SAT-solver. Preliminary experiments show that as the number of levels in feature models increases, our approach manifests a better scalability than the linear-search based approach. This approach can be easily integrated into any feature modeling environment as its verification component.