| --- |
| license: other |
| language: |
| - en |
| tags: |
| - code |
| - ruby |
| - github |
| pretty_name: Ruby Code Corpus |
| size_categories: |
| - 100K<n<1M |
| --- |
| |
| # Ruby Code Corpus |
|
|
| A large corpus of Ruby source code collected from public GitHub repositories. |
|
|
| ## Content |
|
|
| - **294,074 files** from 3,920+ Ruby repositories on GitHub |
| - Sourced from repos ranked by star count (quality signal) |
| - Filtered: removed files under 200 bytes (trivial/empty files) |
| - Only permissive licenses: MIT, Apache-2.0, BSD, ISC, Ruby |
|
|
| ## Fields |
|
|
| | Field | Description | |
| |---|---| |
| | `source` | Always `github` | |
| | `repo` | `owner/repo` slug | |
| | `repo_url` | Full GitHub URL | |
| | `path` | File path within the repo | |
| | `language` | Always `Ruby` | |
| | `license` | SPDX license identifier | |
| | `stars` | GitHub star count at collection time | |
| | `ref` | Branch name | |
| | `size_bytes` | File size in bytes | |
| | `text` | Raw source content | |
|
|
| ## Composition |
|
|
| - ~95% `.rb` files |
| - ~37% test files (spec/test — kept intentionally, useful for training) |
| - ~58% production code |
|
|
| ## Usage |
|
|
| ```python |
| from datasets import load_dataset |
| |
| ds = load_dataset("NickIBrody/ruby-code-corpus", split="train") |
| print(ds[0]["text"]) |
| ``` |
|
|
| ## License |
|
|
| Individual files retain their original open-source licenses as specified in the `license` field. |
|
|
