Paper Details
Abstract
Automation with large language models (LLMs) is transforming software development, including testing one of the most resource-intensive stages of the development life cycle. In this paper, we focus on automating test code generation of behavior-driven development (BDD) scenarios. Currently, from BDD scenarios, one can only generate test function signatures, and the developers have to write test scripts that control the software and verify the results. Given the BDD scenarios and the system under test, we use LLMs to generate end-to-end test code that includes the body of the BDD test functions. This is a significant advancement in automation as fully functional test codes are generated, saving a significant amount of work for developers. The experimental results show great potential for reducing the effort in acceptance testing.