Java Source Code for Program Repair Research

Raw Java source code extracted from GitHub, GitLab, and Bitbucket repositories for training program repair models. It was used in the CoCoNuT research paper and includes commits up to the year 2006. The data has not been shuffled or tokenized.

Use Cases

• Train context-aware neural translation models for automated program repair using raw Java source code sequences.
• Analyze patterns in Java code changes from version control commits up to 2006 for defect prediction.
• Benchmark ensemble learning methods for source code transformation tasks on this specific corpus.

Strengths

• Data is sourced from three major code repositories: GitHub, GitLab, and Bitbucket.
• Specifically created for and used in a published research paper on program repair (CoCoNuT).
• The dataset is in a raw, untokenized state, providing flexibility for preprocessing.

Limitations

• The dataset's temporal coverage is limited, containing only commits up to the year 2006.
• Key structural details like row count, column names, and file size are unknown.
• The license and specific organization of the data are not provided.

Provenance

Source

GitHub, GitLab, and Bitbucket repositories.

Collection Method

Raw data extraction from version control commits.

Time Range

Commits up to the year 2006.

Freshness

The dataset card was last updated on 2023-09-28.