Automated program recognition by graph parsing

Abstract

The recoginition of standard computational structures(clichés) in a program can help an experienced programmerunderstand the program. Based on the known relationshipsbetwwen the clichés, a hierarchical description of theprogram's design can be recovered. We develop and astudy agraph parsing approach to automating program recognition inwhich programs are represented as attributed dataflowgraphs and a library of clichés in the code.We demonstrate that this graph parsing approach is feasibleand useful way to automate program recognition. In studyingthis approach, we have experimented with two medium-sized,real-world simulator programs. There are three aspects ofour study. First, we evaluate our representation's abilityto suppress many common forms of program variation whichhinder recognition. Second, we investigate theexpressiveness of our graph grammar formalism for capturingprogramming clichés. Third, we empirically andanalytically study the computational costs of ourrecognition approach with respect to the real-worldsimulator programs.