Abstract

Program comprehension is an essential part of software maintenance. As software systems are becoming larger, understanding the whole program without prior knowledge is a hard task for the developer and the maintainer. Our approach of feature extraction reduces the program understanding complexity by identifying features based on input and output statements at the source code level. The presented feature extraction approach is a semi-automatic approach that only requires source code and test cases to identify and extract features. This approach utilizes program slicing, a program reduction technique to extract statements that implement an identified feature. The algorithm is implemented within the CONCEPT (C[barbelow]omprehension O[barbelow]f N[barbelow]et- CE ntered P[barbelow]rograms and T[barbelow]echniques) project. A case study using an open source project called JUnit was conducted to evaluate the applicability of the proposed approach.