ParquetDB
Documentation | PyPI | GitHub
ParquetDB is a Python library designed to bridge the gap between traditional file storage and fully fledged databases, all while wrapping the powerful PyArrow library to streamline data input and output. By leveraging the Parquet file format, ParquetDB provides the portability and simplicity of file-based data storage alongside advanced querying features typically found in database systems.
Table of Contents
Documentation
Check out the docs
Features
- Simple Interface: Easy-to-use methods for creating, reading, updating, and deleting data.
- Minimal Overhead: Achieve quick read/write speeds without the complexity of setting up or managing a larger database system.
- Batching: Efficiently handle large datasets by batching operations.
- Supports Complex Data Types: Handles nested and complex data types.
- Schema Evolution: Supports adding new fields and updating schemas seamlessly.
- Supports storing of python objects: ParquetDB can store python objects (objects and functions) using pickle.
- Supports np.ndarrays: ParquetDB can store ndarrays.
Installation
Install ParquetDB using pip:
pip install parquetdb
Usage
Creating a Database
Initialize a ParquetDB instance by specifying the path the name of the dataset
from parquetdb import ParquetDB
db = ParquetDB(db_path='parquetdb')
Adding Data
Add data to the database using the create method. Data can be a dictionary, a list of dictionaries, or a Pandas DataFrame.
data = [
{'name': 'Charlie', 'age': 28, 'occupation': 'Designer'},
{'name': 'Diana', 'age': 32, 'occupation': 'Product Manager'}
]
db.create(data)
Normalizing
Normalization is a crucial process for ensuring the optimal performance and efficient management of data. In the context of file-based databases, like the ones used in ParquetDB, normalization helps balance the distribution of data across multiple files. Without proper normalization, files can end up with uneven row counts, leading to performance bottlenecks during operations like queries, inserts, updates, or deletions.
This method does not return anything but modifies the dataset directory in place, ensuring a more consistent and efficient structure for future operations.
from parquetdb import NormalizeConfig
db.normalize(
normalize_config=NormalizeConfig(
load_format='batches', # Uses the batch generator to normalize
batch_readahead=10, # Controls the number of batches to load in memory a head of time.
fragment_readahead=2, # Controls the number of files to load in memory ahead of time.
batch_size = 100000, # Controls the batchsize when to use when normalizing. This will have impacts on amount of RAM consumed
max_rows_per_file=500000, # Controls the max number of rows per parquet file
max_rows_per_group=500000) # Controls the max number of rows per group parquet file
)
Reading Data
Read data from the database using the read method. You can filter data by IDs, specify columns, and apply filters.
# Read all data
all_employees = db.read()
# Read specific columns
names = db.read(columns=['name'])
# Read data with filters
from pyarrow import compute as pc
age_filter = pc.field('age') > 30
older_employees = db.read(filters=[age_filter])
Updating Data
Update existing records in the database using the update method. Each record must include the id field.
update_data = [
{'id': 1, 'occupation': 'Senior Engineer'},
{'id': 3, 'age': 29}
]
db.update(update_data)
Deleting Data
Delete records from the database by specifying their IDs.
db.delete(ids=[2, 4])
Citing ParquetDB
If you use ParquetDB in your work, please cite the following paper:
@misc{lang2025parquetdblightweightpythonparquetbased,
title={ParquetDB: A Lightweight Python Parquet-Based Database},
author={Logan Lang and Eduardo Hernandez and Kamal Choudhary and Aldo H. Romero},
year={2025},
eprint={2502.05311},
archivePrefix={arXiv},
primaryClass={cs.DB},
url={https://arxiv.org/abs/2502.05311}}
Contributing
Contributions are welcome! Please open an issue or submit a pull request on GitHub. More information can be found in the CONTRIBUTING.md file.
License
This project is licensed under the MIT License. See the LICENSE file for details.
1 maintainer
Metadata
1 maintainer
Release files for parquetdb 1.0.1
For a detailed explanation of source distributions (sdists) and built distributions (wheels), please see the package formats documentation.
Source distribution (sdist)
| File | Size | Uploaded | |
|---|---|---|---|
| parquetdb-1.0.1.tar.gz | 22.8 MB | Details |
Built distribution (wheel)
| File | Interpreter | ABI | Platform | Reset |
|---|---|---|---|---|
| parquetdb-1.0.1-py3-none-any.whl | Python 3 | none | any | Details |
Total release size: 22.8 MB
Release files / parquetdb-1.0.1.tar.gz
| Download URL | parquetdb-1.0.1.tar.gz |
|---|---|
| Size | 22.8 MB |
| Tags | Source |
|
SHA-256 checksum How to use checksums |
04b9fe509993d88c8f8da1d9062522b836e88e52bee83aec54fa420d4adc756e
|
|
BLAKE2b-256 checksum How to use checksums |
a104a89f2d0a9ee5744d0601f856c3b5386361a5265024417ca726291a89198d
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.12.9
|
Release files / parquetdb-1.0.1-py3-none-any.whl
| Download URL | parquetdb-1.0.1-py3-none-any.whl |
|---|---|
| Size | 75.8 kB |
| Tags | Python 3 |
|
SHA-256 checksum How to use checksums |
0f531e906410764391ffd4e66c19439be19dc914f219f698a6d8864e15fd3ad2
|
|
BLAKE2b-256 checksum How to use checksums |
53c203f926bcba70873a735d86f8e0f8cd4e7b642a2933eec9e0b767e05f1daf
|
| Upload date | |
|
Uploaded using Trusted Publishing? What is trusted publishing? |
No |
| Uploaded via |
twine/6.1.0 CPython/3.12.9
|
