| Nome: | Descrição: | Tamanho: | Formato: | |
|---|---|---|---|---|
| 422.25 KB | Adobe PDF |
Orientador(es)
Resumo(s)
Version control is an essential tool in software development, offering numerous well-known advantages, such as maintenance, traceability, collaborative work support, backup, and security. Commits provide the conceptual foundation of version control and contain helpful information for various stakeholders in the development and maintenance of applications, such as date, author, and a brief description of the changes made. Essentially, commits form a chronological sequence of data, but their utility is often underutilized and lacks efficiency in terms of accessibility. This paper presents an architectural solution that enables much greater utilization of commit history,
providing more functional perspectives for accessing this rich source of knowledge—which often conveys experience, practical solutions, the rationale behind those solutions, and even reasons for discarding alternative options. The proposed architecture enhances developers’ comprehension of intricate commit histories by enabling them to examine local changes and extract significant keywords from commit messages and code snippets. Because commit data is presented in a chronological sequence, it can be challenging to track dependencies and contextualize changes (commits). Our method facilitates sophisticated querying and visualization by arranging commit history into a graph structure representing thematic relationships among commits. Components for extracting commit data, processing it via a large language
model (ChatGPT) to identify relevant keywords, storing the outcomes in a graph database, and applying different algorithmic solutions to “redesign” the dependencies between commits are all part of the architecture. The system enables effective keywordbased searching within the integrated development environment by clustering related commits based on shared keywords through community detection algorithms. To better assist developers in managing intricate codebases and enhancing teamwork.
Descrição
Palavras-chave
Git Graphs Knowledge management Large language model Plugins Semantics Soft-ware engineering Version control systems
Contexto Educativo
Citação
Patricio, Manuel; Sousa, Mario; Matos, Paulo; Oliveira, Pedro Filipe (2024). Unstructuring the sequentiality of commits into a semantic network with higher informational and functional quality. In 10th International Conference on Engineering and Emerging Technologies, ICEET 2024. ISSN 2409-2983. p. 1-6
Editora
IEEE