JetBrains-Research/commit-chronicle

📜 CommitChronicle 🔮

This is the dataset for commit message generation (and/or completion), introduced in the paper "From Commit Message Generation to History-Aware Commit Message Completion", ASE 2023.

Its key features:

• large-scale and multilingual: contains 10.7M commits from 11.9k GitHub repositories in 20 programming languages;
• diverse: avoids restrictive filtering on commit messages or commit diffs structure;
• suitable for experiments with commit history: provides metadata about commit authors and dates and uses split-by-project.

Dataset Creation

🔍 For further details, please refer to:

• Paper: https://arxiv.org/abs/2308.07655
• Repository: https://github.com/JetBrains-Research/commit_message_generation

We used GitHub Search tool and official GitHub API to select relevant repositories with permissive licenses (Apache, BSD 3-clause, MIT). On February 9th, 2023, we collected all commits made since 2017 from these repositories via PyDriller. Next, we extensively cleaned the data, including filtering outliers, dropping commits from bot authors, and dropping duplicates. Note: to avoid disclosing personal information, we replaced the commit authors' names and emails with unique identifiers.

Dataset Structure

Data Instances

Each data instance in the dataset is a commit. A commit example would look like the following:

{
  'repo': 'saridormi/commit_chronicle',
  'hash': 'a7fb3b64184f0af5b08285cce14b9139baa94049',
  'author': 123,
  'date': '05.07.2021 15:10:07',
  'timezone': 0,
  'license': 'MIT License',
  'language': 'Jupyter Notebook',
  'message': 'Add license badge to readme',
  'original_message': 'Add license badge to readme',
  'mods': [{'change_type': 'MODIFY',
            'new_path': 'README.md',
            'old_path': 'README.md'
            'diff': '@@ -1,6 +1,6 @@\n'
                    ' # Commits dataset\n'
                    ' \n'
                    '-> :heavy_exclamation_mark: **TODO:** license\n'
                    '+![GitHub](https://img.shields.io/github/license/saridormi/commits_dataset?style=for-the-badge)\n'}],
}

Data Fields

Each example has the following fields:

Field	Description
repo	Commit repository.
hash	Commit hash.
author	Unique id for commit author
date	Commit date (from author).
timezone	Commit timezone (from author).
license	Commit repository's license.
language	Commit repository's main language.
message	Commit message (after processing).
original_message	Commit message (without any processing).
mods	List of file modifications from commit.

Each file modification has the following fields:

Field	Description
change_type	Type of change to current file. One of: ADD, COPY, RENAME, DELETE, MODIFY or UNKNOWN.
old_path	Path to file before change (might be empty).
new_path	Path to file after change (might be empty).
diff	git diff for current file.

Data Splits

We provide the following configurations:

• default
  • train: full training split (7.66M commits)
  • validation: full validation split (1.55M commits)
  • test: full test split (1.49M commits)
• subset_cmg
  • test: test subset used for experiments with CMG approaches (204k commits)
• subset_llm
  • test: test subset used for experiments with a LLM (4k commits)

Considerations for Using the Data

Adopted from the Stack.

The released dataset may contain sensitive information such as emails, IP addresses, and API/ssh keys that have previously been published to public repositories on GitHub. In the event that the dataset contains personal information, researchers should only use public, non-personal information in support of conducting and publishing their open-access research. Personal information should not be used for spamming purposes, including sending unsolicited emails or selling of personal information.

The dataset is a collection of commits from repositories with various licenses. Any use of all or part of the code gathered in this dataset must abide by the terms of the original licenses, including attribution clauses when relevant. We facilitate this by providing provenance information for each data point.

Citation

TODO