Real Coders, Real Feedback: Unveiling GitHub Copilot’s Impact on Test Generation Through Practitioner Insights

Nome do aluno

 

Cristiano dos Santos e Santos

 

Título do trabalho

 

Real Coders, Real Feedback: Unveiling GitHub Copilot’s Impact on Test Generation Through Practitioner Insights

 

Resumo do trabalho

 

Automated test generation tools play a key role in software development, yet concerns persist regarding their readability, adoption in industry, and effectiveness in real-world scenarios. Recent studies have employed Large Language Models (LLMs) for test generation, with GitHub Copilot emerging as a prominent tool leveraging LLMs for code suggestions. However, empirical investigations into Copilot’s utility in real-world development scenarios still need to be available. This paper minimizes this gap by conducting an empirical study to leverage the challenges practitioners face when using GitHub Copilot for test generation. Two research directions guide the investigation: (i) identifying anomalies in unit tests generated by Copilot and (ii) exploring practitioners’ perceptions of Copilot’s impact on their productivity in the testing process. The study involved 12 Software Engineers, who gave us their impressions of using GitHub Copilot in practice. The yielded results reveal challenges in Copilot-generated tests, requiring substantial human intervention for compilation correctness, handling testing mocks, inferring parameter types, and needing more standards to handle exception testing. Participants’ reliance on heavy intervention to fix Copilot’s suggestions indicates a need for more trust and refinement in the tool’s output. Additionally, the study acknowledges the limitations of conducting empirical research within a specific context, influencing Copilot’s behavior and highlighting the need for broader studies that encompass other practitioners’ perspectives.

 

Orientador

 

Ivan do Carmo Machado

 

Co-orientador

 

Márcio de Medeiros Ribeiro

 

Membro Titular 1

 

Cláudio Nogueira Sant'Anna

 

Link para o curriculum lattes

 

http://lattes.cnpq.br/3228159608138969

 

Membro Titular 2

 

Rohit Gheyi

 

Link para o curriculum lattes

 

http://lattes.cnpq.br/2931270888717344

 

Suplente 1

 

Carla Ilane Moreira Bezerra

 

Link para o curriculum lattes

 

http://lattes.cnpq.br/4277471687235814

 

Suplente 2

 

Larissa Rocha Soares Bastos

 

Link para o curriculum lattes

 

http://lattes.cnpq.br/5750570352089990

 

Data do exame

 

27 Aug, 2024

 

Horário do exame

 

9:00 AM

 

 

Data da Defesa: 
27/08/2024 - 09:00
Tipo de Defesa: 
Qualificação de Mestrado