Encoding Version History Context for Better Code Representation

TL;DR

This work addresses the gap that current deep learning models for code primarily rely on source code, neglecting richer contextual artefacts from software evolution. The authors encode version history, call hierarchy, and temporal cues into existing architectures (ASTNN and CodeBERT) and evaluate on code clone detection and code classification. They report improvements in both tasks when adding version history and other artefacts, with significant gains under certain aggregations and model choices, but also highlight limitations like token-length constraints and dataset size. The study outlines a concrete research agenda to systematically investigate how to best encode diverse software contexts to enhance code understanding in real-world, larger-scale settings.

Abstract

With the exponential growth of AI tools that generate source code, understanding software has become crucial. When developers comprehend a program, they may refer to additional contexts to look for information, e.g. program documentation or historical code versions. Therefore, we argue that encoding this additional contextual information could also benefit code representation for deep learning. Recent papers incorporate contextual data (e.g. call hierarchy) into vector representation to address program comprehension problems. This motivates further studies to explore additional contexts, such as version history, to enhance models' understanding of programs. That is, insights from version history enable recognition of patterns in code evolution over time, recurring issues, and the effectiveness of past solutions. Our paper presents preliminary evidence of the potential benefit of encoding contextual information from the version history to predict code clones and perform code classification. We experiment with two representative deep learning models, ASTNN and CodeBERT, to investigate whether combining additional contexts with different aggregations may benefit downstream activities. The experimental result affirms the positive impact of combining version history into source code representation in all scenarios; however, to ensure the technique performs consistently, we need to conduct a holistic investigation on a larger code base using different combinations of contexts, aggregation, and models. Therefore, we propose a research agenda aimed at exploring various aspects of encoding additional context to improve code representation and its optimal utilisation in specific situations.

Figures (1)

• Figure 1: A motivating example of using Version History to detect code clones.