Applying program analysis to code retrieval

Abstract

Early code retrieval systems were primarily adaptations of standard text retrieval approaches, and so treated source code as either plain or structured text. While fairly successful, these approaches ignored much of the information that can be extracted from the source code. Recently, researchers have demonstrated a number of ways in which static program analysis can be used to augment text-based retrieval approaches. By taking advantage of the structural and semantic information embedded in source code, advanced code retrieval systems can provide a superior experience. This chapter begins describing how basic text-based code retrieval systems function. It then introduces a basic form of static program analysis which allows source code to be treated as structured text. Finally, it describes link analysis, an advanced program analysis technique. Link analysis aids code retrieval systems in numerous ways, for example enabling better estimates of result quality and the sharing of descriptive terms. The chapter concludes by describing in great detail a single static program analysis technique called dependency slicing. Dependency slicing is used in code retrieval systems to package up search results as a compilable unit, which supports the reuse of the retrieved results.