Commit Chronicle: 10.7 Million Multilingual GitHub Commits for Message Generation

CommitChronicle is a large-scale dataset for commit message generation and completion, introduced in an ASE 2023 paper. It contains 10.7 million commits from 11.9 thousand GitHub repositories across 20 programming languages. The dataset was created by JetBrains-Research and last updated on Hugging Face in October 2023.

Use Cases

• Training models for automated commit message generation based on commit diffs.
• Experimenting with history-aware commit message completion using provided metadata.
• Analyzing patterns in commit messages across 20 programming languages.
• Studying the diversity of commit structures and messages in large-scale open-source projects.

Strengths

• Large scale with 10.7 million commits.
• Multilingual coverage from 11.9 thousand repositories in 20 programming languages.
• Diverse content due to minimal filtering on commit messages and diffs.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count is unknown, which may limit suitability assessment.
• Last updated 2023-10-05 10:50:00; freshness should be verified.

Provenance

Source

GitHub repositories.

Collection Method

Likely collected via GitHub API or similar methods, as described in the associated ASE 2023 paper.

Time Range

null

Freshness

2023-10-05

Geography

null