Temporal analysis on pull request patterns: an approach with sliding window

• Silvana Gonçalves IFAC
• Daricélio Soares UFAC
• Daniel Silva UFAC

Resumo

In open-source software projects, it is common for external collaborators to send contributions through pull requests, with bug fixes, new features and/or any type of modification in software artifacts. These contributions are evaluated by the core team that decides on their acceptance. This pull requests evaluation process has been the object of study by researchers who seek to understand which factors influence this stage. Some works used association rules to show patterns that indicate the influencing factors in various scenarios, namely: acceptance, lifetime, and assignment of reviewers. However, in general, these results do not consider the time perspective inherent in the collaborative process and software evolution. The temporal analysis of patterns on pull requests, through association rules, is configured as an open research topic. This analysis is important, above all, for development teams, so that they can identify the moment when certain patterns have occurred with greater intensity, allowing a contextual and qualitative assessment of the projects at a given moment. In addition, a more detailed assessment can allow development teams to react appropriately to the patterns found, enhancing practices that lead to positive patterns and/or avoiding mistaken actions to mitigate undesired patterns. The goal of this work is to propose an approach of temporal analysis of association rules on pull requests that allows a more detailed interpretation of the extracted patterns. The results found show patterns that: i) gain/lose strength over time, regardless of the partitioning used; ii) do not exist for some periods of time; and iii) show a tendency to stability.

Palavras-chave: association rules, Pull request, temporal analysis