An approach to retrieving similar source codes by control structure and method identifiers

Abstract

In this paper, we present an approach to improve source code retrieval using the structure of control statements. We develop a lexical parser that extracts structural information for each method including control statements and method identifiers. We present a source code retrieval model, named "derived structure retrieval model," in which a retrieval conditions are defined as a sequence of control statements and/or method identifiers, and meaningful search conditions are derived from the given sequence. Experiments on the source code of Struts 2 Core show that the derived structure retrieval model outperforms the vector space retrieval model. © 2013 Springer-Verlag.