pretty_name: Python–Java Project Dataset
language:
- en
tags:
- code
- code-completion
- program-repair
- repository-level
- python
- java
size_categories:
- n<1K
configs:
- config_name: python
data_files:
- split: test
path: python.jsonl
- config_name: java
data_files:
- split: test
path: java.jsonl
Python–Java Project Dataset
This dataset provides metadata, original GitHub repository links, and pinned commit identifiers for 100 projects: 50 Python projects and 50 Java projects.
These small-scale projects originate from real-world GitHub repositories and span a range of application domains. Their original repository structures provide software contexts beyond isolated programming exercises.
Subsets and access
Select python or java in the dataset viewer. Each subset contains 50 project records with the same schema. Download individual metadata files through the Files and versions tab. Upstream source archives are linked in each record; they are hosted by the original repository owners.
With the Hugging Face datasets package, use the following with this dataset's repository identifier:
from datasets import load_dataset
python_projects = load_dataset("gracezheng72/python-java-project-dataset", "python", split="test", revision="1349e5177c968f789938ff50d1daf6f787b19d87")
java_projects = load_dataset("gracezheng72/python-java-project-dataset", "java", split="test", revision="1349e5177c968f789938ff50d1daf6f787b19d87")
Provenance and local verification
The final language-specific metadata CSVs were compared with the local experimental copies and matched in all 100 records. Git checkouts were checked against their recorded commit and tracked-file state. Projects originally downloaded without Git history were checked against their saved file-hash copy manifests, including comparison with the desktop originals. See provenance_audit.json for counts and qualifications.
Older root-level download manifests contain replaced candidates and are not the final cohort definition. The original language-ratio values are preserved as recorded metadata; they were not independently recomputed during release preparation.
Fields
| Field | Meaning |
|---|---|
project_id |
Stable language-prefixed project identifier |
programming_language |
Python or Java |
repository, github_url |
Upstream owner/name and repository link |
commit_sha, commit_url |
Exact recorded source revision and its GitHub link |
source_archive_url |
ZIP link for that upstream commit |
project_directory |
Original experiment directory identifier |
dataset_status |
Status in the final project metadata |
recorded_primary_language_ratio |
Language ratio recorded in the original metadata |
provenance_verification |
Local checks supporting this record |
root_license_files |
License filenames observed at the local project root; not a definitive license classification |
source_redistributed_in_this_release |
False for this metadata-only release |
license_note |
Upstream-license qualification |
Attribution and licenses
All upstream project code remains attributed to its original authors and subject to its own applicable license and notices. A public GitHub repository does not by itself establish permission to redistribute its code. Some projects require additional license verification; therefore no upstream source code is republished in this initial release and no single license is claimed for the collected project code. Follow each pinned repository's license when retrieving or using source files.
Limitations and reproducibility
Pinned source links identify revisions but do not guarantee continued upstream availability. Remote ZIP availability was not rechecked in the local audit. GitHub source archives omit Git history and may omit submodule contents or contain Git LFS pointers; obtain any required submodule/LFS data from the upstream project's instructions. Full replication additionally requires the exact skeleton policy, prompts, protected tests, environment configuration, and evaluation scripts, which are outside this index release.
No model-ranking, pass-rate, security, or contamination conclusions should be inferred from these repository indexes. Project order and metadata do not establish freedom from training-data overlap.
Citation and versioning
When citing this index, include the dataset page URL and its release tag or commit SHA. The associated manuscript is under revision; no publication venue, DOI, or finalized paper citation is asserted here.