Datasets:
Update README.md
Browse files
README.md
CHANGED
|
@@ -65,49 +65,55 @@ size_categories:
|
|
| 65 |
- 10K<n<100K
|
| 66 |
---
|
| 67 |
|
| 68 |
-
# GitHub
|
| 69 |
|
| 70 |
-
A **curated, high-signal dataset
|
| 71 |
-
Each
|
| 72 |
|
| 73 |
-
|
| 74 |
-
|
| 75 |
-
- **
|
| 76 |
-
- **
|
| 77 |
-
- **
|
|
|
|
| 78 |
|
| 79 |
---
|
| 80 |
|
| 81 |
-
##
|
| 82 |
|
| 83 |
-
|
| 84 |
|
| 85 |
To maintain quality and usefulness:
|
| 86 |
-
- Only **closed issues** were considered
|
| 87 |
-
- Each issue must have a **clearly associated fix**
|
| 88 |
-
- Fixes are stored as **unified diffs** extracted from the resolving commit
|
| 89 |
-
- Low-signal issues (questions, duplicates, discussions) were filtered out
|
| 90 |
-
- Issues without meaningful code changes were excluded
|
| 91 |
|
| 92 |
-
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 93 |
|
| 94 |
---
|
| 95 |
|
| 96 |
-
## Dataset
|
| 97 |
|
| 98 |
-
Each
|
| 99 |
|
| 100 |
```json
|
| 101 |
{
|
| 102 |
"repo": "owner/repository",
|
| 103 |
-
"
|
| 104 |
-
"
|
| 105 |
-
"
|
| 106 |
-
"
|
|
|
|
|
|
|
|
|
|
| 107 |
"files": [
|
| 108 |
{
|
| 109 |
"filename": "path/to/file.ext",
|
| 110 |
-
"patch": "unified diff
|
| 111 |
"additions": 10,
|
| 112 |
"deletions": 2
|
| 113 |
}
|
|
@@ -115,19 +121,21 @@ Each dataset entry has the following schema:
|
|
| 115 |
}
|
| 116 |
```
|
| 117 |
|
| 118 |
-
| Field | Description
|
| 119 |
-
| ------------------- | -------------------------------------------- |
|
| 120 |
-
| `repo` | GitHub repository
|
| 121 |
-
| `
|
| 122 |
-
| `
|
| 123 |
-
| `
|
| 124 |
-
| `
|
| 125 |
-
| `
|
| 126 |
-
| `
|
| 127 |
-
| `
|
| 128 |
-
| `files
|
| 129 |
-
| `files[].
|
| 130 |
-
|
|
|
|
|
|
|
| 131 |
|
| 132 |
## Supported languages
|
| 133 |
|
|
@@ -147,22 +155,23 @@ Language distribution varies by repository.
|
|
| 147 |
|
| 148 |
This dataset is well-suited for:
|
| 149 |
|
| 150 |
-
* Training models to generate
|
| 151 |
-
*
|
| 152 |
-
* Building autonomous debugging or
|
| 153 |
-
* Research
|
| 154 |
|
| 155 |
-
It is
|
| 156 |
|
| 157 |
-
*
|
| 158 |
-
*
|
| 159 |
-
*
|
| 160 |
|
| 161 |
## Limitations
|
| 162 |
-
|
| 163 |
-
|
| 164 |
-
|
| 165 |
-
|
| 166 |
-
|
| 167 |
-
|
| 168 |
-
|
|
|
|
|
|
| 65 |
- 10K<n<100K
|
| 66 |
---
|
| 67 |
|
| 68 |
+
# GitHub Pull Request Bug–Fix Dataset
|
| 69 |
|
| 70 |
+
A **curated, high-signal dataset of real-world software bugs and fixes** collected from **25 popular open-source GitHub repositories**.
|
| 71 |
+
Each entry corresponds to a **single pull request (PR)** and pairs contextual metadata with the **exact code changes (unified diffs)** that fixed the bug.
|
| 72 |
|
| 73 |
+
This dataset is designed for:
|
| 74 |
+
|
| 75 |
+
- **Automated program repair**
|
| 76 |
+
- **Bug-fix patch generation**
|
| 77 |
+
- **LLM-based code and debugging agents**
|
| 78 |
+
- **Empirical software engineering research**
|
| 79 |
|
| 80 |
---
|
| 81 |
|
| 82 |
+
## Data collection methodology
|
| 83 |
|
| 84 |
+
Data was collected using the **GitHub REST API** and post-processed into a structured format.
|
| 85 |
|
| 86 |
To maintain quality and usefulness:
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
| 87 |
|
| 88 |
+
- Only **merged pull requests** were included
|
| 89 |
+
- Each PR must represent a **bug fix or correctness change**
|
| 90 |
+
- Fixes are linked to both the **buggy commit** and the **fix commit**
|
| 91 |
+
- Code changes are stored as **unified diffs** at the file level
|
| 92 |
+
- Low-signal PRs (refactors, formatting-only changes, discussions) were filtered out
|
| 93 |
+
- PRs without meaningful code changes were excluded
|
| 94 |
+
|
| 95 |
+
Each dataset row represents **one bug–fix PR**.
|
| 96 |
|
| 97 |
---
|
| 98 |
|
| 99 |
+
## Dataset schema
|
| 100 |
|
| 101 |
+
Each entry follows the schema below:
|
| 102 |
|
| 103 |
```json
|
| 104 |
{
|
| 105 |
"repo": "owner/repository",
|
| 106 |
+
"pr_number": 12345,
|
| 107 |
+
"title": "Short description of the fix",
|
| 108 |
+
"body": "Pull request description and context",
|
| 109 |
+
"buggy_commit": "abcdef123456...",
|
| 110 |
+
"fix_commit": "fedcba654321...",
|
| 111 |
+
"buggy_distance": 12,
|
| 112 |
+
"confidence": "medium",
|
| 113 |
"files": [
|
| 114 |
{
|
| 115 |
"filename": "path/to/file.ext",
|
| 116 |
+
"patch": "unified diff representing the fix",
|
| 117 |
"additions": 10,
|
| 118 |
"deletions": 2
|
| 119 |
}
|
|
|
|
| 121 |
}
|
| 122 |
```
|
| 123 |
|
| 124 |
+
| Field | Description |
|
| 125 |
+
| ------------------- | ------------------------------------------------- |
|
| 126 |
+
| `repo` | GitHub repository containing the pull request |
|
| 127 |
+
| `pr_number` | Pull request number |
|
| 128 |
+
| `title` | Pull request title |
|
| 129 |
+
| `body` | Pull request description and discussion context |
|
| 130 |
+
| `buggy_commit` | Commit introducing or containing the bug |
|
| 131 |
+
| `fix_commit` | Commit that fixes the bug |
|
| 132 |
+
| `buggy_distance` | Number of commits between buggy and fix commits |
|
| 133 |
+
| `confidence` | Heuristic confidence level of bug–fix correctness |
|
| 134 |
+
| `files` | List of files modified by the fix |
|
| 135 |
+
| `files[].filename` | Path to the modified file |
|
| 136 |
+
| `files[].patch` | Unified diff containing the code changes |
|
| 137 |
+
| `files[].additions` | Number of lines added |
|
| 138 |
+
| `files[].deletions` | Number of lines removed |
|
| 139 |
|
| 140 |
## Supported languages
|
| 141 |
|
|
|
|
| 155 |
|
| 156 |
This dataset is well-suited for:
|
| 157 |
|
| 158 |
+
* Training models to generate patches from real pull request context
|
| 159 |
+
* Studying bug-fix patterns across large codebases
|
| 160 |
+
* Building autonomous debugging or repair agents
|
| 161 |
+
* Research in program repair, code synthesis, and software maintenance
|
| 162 |
|
| 163 |
+
It is not intended for:
|
| 164 |
|
| 165 |
+
* Pull request classification or triage
|
| 166 |
+
* Sentiment analysis
|
| 167 |
+
* General-purpose chatbot fine-tuning without code generation objectives
|
| 168 |
|
| 169 |
## Limitations
|
| 170 |
+
The dataset reflects real-world noise from GitHub pull requests
|
| 171 |
+
Buggy commit identification is heuristic and may be imperfect
|
| 172 |
+
Some fixes involve refactoring or design changes rather than minimal patches
|
| 173 |
+
No guarantee that fixes represent optimal or best-practice solutions
|
| 174 |
+
|
| 175 |
+
> ⚠️ Warning
|
| 176 |
+
This dataset currently includes pull requests from 10 of the planned 25 repositories, totaling approximately 24,000 entries.
|
| 177 |
+
The final release is expected to contain ~50,000 entries and occupy ~2 GB of storage.
|