SAGA: Summarization-Guided Assert Statement Generation

内容摘要

Generating meaningful assert statements is one of the key challenges in automated test case generation, which requires understanding the intended functionality of the tested code. Recently, deep learning-based models have shown promise in improving the performance of assert statement generation. However, existing models only rely on the test prefixes along with their corresponding focal methods, yet ignore the developer-written summarization. Based on our observations, the summarization contents usually express the intended program behavior or contain parameters that will appear directly in the assert statement. Such information will help existing models address their current inability to accurately predict assert statements. This paper presents a novel summarization-guided approach for automatically generating assert statements. To derive generic representations for natural language (i.e., summarization) and programming language (i.e., test prefixes and focal methods), we leverage a pre-trained language model as the reference architecture and fine-tune it on the task of assert statement generation. To the best of our knowledge, the proposed approach makes the first attempt to leve