Measuring Unique Changes: How do Distinct Changes Affect the Size and Lifetime of Pull Requests?
Resumo
Size metrics are commonly cited features in studies that analyze influencing factors on pull request lifetime. These metrics are also important for integrators, as based on them, they may prefer to prioritize pull requests easy to assess. However, code changes that form pull requests may not be unique, and repetitive changes may represent less complexity than expected by considering size metrics like lines of code or source files. The goal of this study is to analyze the influence of unique changes over pull requests relative to its size and lifetime. We collected data from 83,000+ pull requests of 26 projects hosted on GitHub. Also, we proposed a metric called unique changes rate to measure the proportion of unique changes over the total changes made by a pull request. We conducted experiments with Random Forest regression models and association rules to examine the influence of unique changes rate. Results show that unique changes have more influence over the lifetime of large pull requests, is determined mainly by the number of source files, and low levels of unique changes rate affect more the relationship between pull request size and lifetime than high levels. We conclude that unique changes can figure as an interesting feature in the context of the pull request lifetime. Results indicate that unique changes may increase or decrease the influence of pull request size on its lifetime. Our work has implications for researchers and core team members in software projects since unique changes represent helpful information.
Palavras-chave:
Pull request, association rules, lifetime, unique changes
Publicado
19/10/2020
Como Citar
SILVA, Daniel Augusto Nunes da; SOARES, Daricélio Moreira; GONÇALVES, Silvana Andrade.
Measuring Unique Changes: How do Distinct Changes Affect the Size and Lifetime of Pull Requests?. In: SIMPÓSIO BRASILEIRO DE COMPONENTES, ARQUITETURAS E REUTILIZAÇÃO DE SOFTWARE (SBCARS), 14. , 2020, Natal/RN.
Anais [...].
Porto Alegre: Sociedade Brasileira de Computação,
2020
.
p. 121–130.