Code Representation Pre-training with Complements from Program Executions

TL;DR

This paper introduces FuzzPretrain, a pre-training method that incorporates dynamic execution information from test cases to enhance code representations, leading to improved code search performance.

Contribution

It proposes a novel pre-training approach that leverages program execution data from test cases to improve code understanding beyond syntactic features.

Findings

FuzzPretrain improves code search mAP by over 6% and 9% compared to source code and AST-based models.

Learning from program executions provides more discriminative code representations.

Experimental results validate the effectiveness of incorporating dynamic execution information.

Abstract

Large language models (LLMs) for natural language processing have been grafted onto programming language modeling for advancing code intelligence. Although it can be represented in the text format, code is syntactically more rigorous in order to be properly compiled or interpreted to perform a desired set of behaviors given any inputs. In this case, existing works benefit from syntactic representations to learn from code less ambiguously in the forms of abstract syntax tree, control-flow graph, etc. However, programs with the same purpose can be implemented in various ways showing different syntactic representations while the ones with similar implementations can have distinct behaviors. Though trivially demonstrated during executions, such semantics about functionality are challenging to be learned directly from code, especially in an unsupervised manner. Hence, in this paper, we propose FuzzPretrain to explore the dynamic information of programs revealed by their test cases and embed it into the feature representations of code as complements. The test cases are obtained with the assistance of a customized fuzzer and are only required during pre-training. FuzzPretrain yielded more than 6%/9% mAP improvements on code search over its counterparts trained with only source code or AST, respectively. Our extensive experimental results show the benefits of learning discriminative code representations with program executions.