helmo/github-issues

HuggingFace Datasets Repository Issues

Dataset Description

This dataset contains issues and pull requests from the huggingface/datasets repository, collected via the GitHub API. Each entry includes comprehensive metadata about the issue/PR along with all associated comments, making it valuable for studying software development patterns, issue resolution processes, and community interactions in open-source projects.

Dataset Summary

• Repository: huggingface/datasets
• Total Issues/PRs: 7,540
• Date Collected: June 13, 2025
• Language: English
• License: Apache 2.0

The dataset includes both open and closed issues/pull requests with their complete comment threads, providing rich context for understanding how software issues are discussed and resolved in a major open-source machine learning library.

Supported Tasks and Leaderboards

This dataset can be used for various NLP and software engineering research tasks:

• Text Classification: Categorizing issues by type (bug, feature request, question, etc.)
• Sentiment Analysis: Analyzing the tone of issue discussions
• Text Generation: Generating responses to software issues
• Question Answering: Extracting answers from issue discussions
• Software Engineering Research: Studying issue resolution patterns, community interactions, and development workflows

Languages

The dataset is primarily in English, as it contains issues from an English-speaking open-source community.

Dataset Structure

Data Instances

Each instance represents a single GitHub issue or pull request with the following structure:

{
  "number": 7613,
  "title": "fix parallel push_to_hub in dataset_dict",
  "body": "Description of the issue...",
  "state": "open",
  "user": {
    "login": "username",
    "id": 12345,
    ...
  },
  "labels": [
    {
      "name": "bug",
      "color": "d73a4a",
      ...
    }
  ],
  "comments": [
    "First comment text...",
    "Second comment text...",
    ...
  ],
  "created_at": "2025-06-13T09:02:24Z",
  "updated_at": "2025-06-13T10:38:04Z",
  "pull_request": {
    "url": "https://api.github.com/repos/huggingface/datasets/pulls/7613",
    ...
  },
  ...
}

Data Fields

• number (int64): Issue/PR number
• title (string): Title of the issue/PR
• body (string): Main description/content
• state (string): Current state (open/closed)
• user (struct): Information about the user who created the issue
• labels (list): Labels assigned to the issue
• comments (sequence): List of all comment texts
• created_at (timestamp): Creation timestamp
• updated_at (timestamp): Last update timestamp
• closed_at (timestamp): Closure timestamp (if closed)
• pull_request (struct): PR-specific metadata (if applicable)
• assignee/assignees (struct/list): Assigned users
• milestone (struct): Associated milestone information
• reactions (struct): Reaction counts (+1, -1, etc.)
• author_association (string): Relationship to repository (OWNER, CONTRIBUTOR, etc.)

Data Splits

The dataset contains a single split:

• train: 7,540 issues/pull requests

Dataset Creation

Curation Rationale

This dataset was created to provide researchers and developers with real-world examples of software issue discussions and resolutions from a popular machine learning library. It can help understand:

• How technical issues are communicated and resolved
• Patterns in community interaction and support
• Evolution of software projects through issue tracking
• Natural language patterns in technical documentation

Source Data

Initial Data Collection and Normalization

The data was collected using the GitHub API v4, specifically targeting the huggingface/datasets repository. The collection process:

1. Issues Retrieval: All issues and pull requests were fetched using paginated API calls
2. Comments Collection: For each issue/PR, all associated comments were retrieved
3. Data Processing: The raw JSON responses were processed and structured into a consistent format
4. Timestamp Handling: All timestamps were normalized to UTC format

Who are the source language producers?

The language producers are contributors to the HuggingFace datasets library, including:

• HuggingFace team members and maintainers
• Open-source contributors from the global developer community
• Users reporting bugs and requesting features
• Community members providing support and discussions

Annotations

Annotation process

No additional annotations were added beyond the existing GitHub metadata (labels, assignees, milestones, etc.) that were already present in the repository.

Who are the annotators?

The repository maintainers and contributors who applied labels and other metadata during normal issue management processes.

Personal and Sensitive Information

The dataset contains publicly available information from GitHub issues. While no intentionally sensitive information should be present, users should be aware that:

• GitHub usernames and profile information are included
• Some issues might contain system information, file paths, or configuration details
• Email addresses might appear in code snippets or error messages

Considerations for Using the Data

Social Impact of Dataset

This dataset can contribute positively to software engineering research and education by providing insights into collaborative development processes. However, users should consider:

• Privacy: Respect the public nature of the data and avoid any analysis that could harm individual contributors
• Context: Issues represent specific technical problems and may not generalize to all software projects
• Bias: The dataset reflects the specific community and practices of the HuggingFace ecosystem

Discussion of Biases

Potential biases in the dataset include:

• Language Bias: Primarily English-language content
• Domain Bias: Focused on machine learning/data science library issues
• Community Bias: Reflects the practices and communication style of the HuggingFace community
• Temporal Bias: Represents issues from a specific time period in the project's evolution
• Technical Bias: May over-represent certain types of technical issues common in ML libraries

Other Known Limitations

• The dataset represents a snapshot from June 13, 2025, and doesn't include subsequent updates
• Comment threads are included as lists but don't preserve detailed threading structure
• Some metadata fields may be incomplete for older issues
• The dataset doesn't include private repository discussions or communications

Additional Information

Dataset Curators

This dataset was curated by Hélder Monteiro by extracting and processing public information from the HuggingFace datasets repository using the GitHub API.

Licensing Information

This dataset is licensed under the Apache 2.0 License, consistent with the open-source nature of the original repository.

Citation Information

@dataset{monteiro_huggingface_datasets_issues_2025,
  title={HuggingFace Datasets Repository Issues},
  author={Hélder Monteiro},
  year={2025},
  month={June},
  url={https://huggingface.co/datasets/helmo/github-issues},
  note={Issues and pull requests from huggingface/datasets repository collected and curated via GitHub API}
}

Disclaimer

This dataset contains content created by the HuggingFace community members who opened issues, submitted pull requests, and participated in discussions on the datasets repository. While this dataset compilation is provided as it is, all original content belongs to the respective contributors.

For questions about this dataset or to report issues, please open an issue in the dataset repository or contact the dataset maintainers.