| | --- |
| | pretty_name: SchemaPile |
| | tags: |
| | - SQL |
| | - Schemas |
| | dataset_info: |
| | features: |
| | - name: ID |
| | dtype: string |
| | - name: FILENAME |
| | dtype: string |
| | - name: URL |
| | dtype: string |
| | - name: LICENSE |
| | dtype: string |
| | - name: PERMISSIVE |
| | dtype: bool |
| | - name: TABLES |
| | list: |
| | - name: TABLE_NAME |
| | dtype: string |
| | - name: COLUMNS |
| | list: |
| | - name: NAME |
| | dtype: string |
| | - name: TYPE |
| | dtype: string |
| | - name: NULLABLE |
| | dtype: bool |
| | - name: UNIQUE |
| | dtype: bool |
| | - name: DEFAULT |
| | dtype: string |
| | - name: CHECKS |
| | sequence: string |
| | - name: IS_PRIMARY |
| | dtype: bool |
| | - name: IS_INDEX |
| | dtype: bool |
| | - name: VALUES |
| | sequence: string |
| | - name: PRIMARY_KEYS |
| | sequence: string |
| | - name: FOREIGN_KEYS |
| | list: |
| | - name: COLUMNS |
| | sequence: string |
| | - name: FOREIGN_TABLE |
| | dtype: string |
| | - name: REFERRED_COLUMNS |
| | sequence: string |
| | - name: ON_DELETE |
| | dtype: string |
| | - name: ON_UPDATE |
| | dtype: string |
| | - name: CHECKS |
| | sequence: string |
| | - name: INDEXES |
| | sequence: string |
| | splits: |
| | - name: full |
| | num_bytes: 142346554 |
| | num_examples: 22989 |
| | download_size: 39308883 |
| | dataset_size: 142346554 |
| | configs: |
| | - config_name: default |
| | data_files: |
| | - split: full |
| | path: data/full-* |
| | --- |
| | |
| | # SchemaPile |
| |
|
| | ### Usage |
| |
|
| | ```python |
| | from datasets import load_dataset |
| | ds = load_dataset("trl-lab/schemapile", split="full") |
| | print(f"Loaded dataset with {len(ds)} records.") |
| | print(ds[0]) |
| | ``` |
| |
|
| | ### Description |
| |
|
| | SchemaPile is a collection of database schemas extracted from various sources, normalized for consistency and ease of use in machine learning workflows. Each record contains metadata (`INFO`), licensing information, permissiveness, and a list of tables with detailed column specifications. |
| |
|
| | ### Schema |
| |
|
| | - **INFO:** |
| | - `ID`: Unique identifier (from filename prefix) |
| | - `FILENAME`: Original filename |
| | - Additional metadata fields |
| |
|
| | - **LICENSE:** License string (if available) |
| | - **PERMISSIVE:** Boolean indicating permissive license |
| | - **TABLES:** List of tables, each with: |
| | - `TABLE_NAME`: Name of the table |
| | - `COLUMNS`: List of columns, each with: |
| | - `NAME`, `TYPE`, `NULLABLE`, `UNIQUE`, `DEFAULT`, `CHECKS`, `IS_PRIMARY`, `IS_INDEX`, `VALUES` |
| | - `PRIMARY_KEYS`, `FOREIGN_KEYS`, `CHECKS`, `INDEXES` |
| |
|
| | ### Citation |
| |
|
| | If you use SchemaPile in your research, please cite: |
| |
|
| | > ```bibtex |
| | > @article{dohmen2024schemapile, |
| | > title={Schemapile: A large collection of relational database schemas}, |
| | > author={D{"o}hmen, Till and Geacu, Radu and Hulsebos, Madelon and Schelter, Sebastian}, |
| | > journal={Proceedings of the ACM on Management of Data}, |
| | > volume={2}, |
| | > number={3}, |
| | > pages={1--25}, |
| | > year={2024}, |
| | > publisher={ACM New York, NY, USA} |
| | > ``` |
| |
|
| | **[Read the paper](https://schemapile.github.io/assets/SchemaPile.pdf)** |
| |
|
| | ### Website |
| |
|
| | For more information, visit the [SchemaPile website](https://schemapile.github.io/). |
| |
|
| | ### License |
| |
|
| | See individual record `LICENSE` fields for details. The `PERMISSIVE` flag indicates whether the schema is under a permissive license. |
| |
|
