Search-based amorphous slicing

Abstract

Amorphous slicing is an automated source code extraction technique with applications in many areas of software engineering, including comprehension, reuse, testing and reverse engineering. Algorithms for syntax-preserving slicing are well established, but amorphous slicing is harder because it requires arbitrary transformation; finding good general purpose amorphous slicing algorithms therefore remains as hard as general program transformation. In this paper we show how amorphous slices can be computed using search techniques. The paper presents results from a set of experiments designed to explore the application of genetic algorithms, hill climbing, random search and systematic search to a set of six subject programs. As a benchmark, the results are compared to those from an existing analytical algorithm for amorphous slicing, which was written specifically to perform well with the sorts of program under consideration. The results, while tentative at this stage, do give grounds for optimism. The search techniques proved able to reduce the size of the programs under consideration in all cases, sometimes equaling the performance of the specifically-tailored analytic algorithm. In one case, the search techniques performed better, highlighting a fault in the existing algorithm