Kùzu is an in-process property graph database management system (GDBMS) built for query speed and scalability. Kùzu is optimized for handling complex join-heavy analytical workloads on very large graph databases, with the following core feature set:
- Flexible Property Graph Data Model and Cypher query language
- Embeddable, serverless integration into applications
- Columnar disk-based storage
- Columnar sparse row-based (CSR) adjacency list/join indices
- Vectorized and factorized query processor
- Novel and very fast join algorithms
- Multi-core query parallelism
- Serializable ACID transactions
Kùzu is being actively developed at University of Waterloo as a feature-rich and usable GDBMS. Kùzu is available under a permissible license. So try it out and help us make it better! We welcome your feedback and feature requests.
To build from source code, Kùzu requires Cmake(>=3.11), Python 3, and a compiler that supports C++20.
- Perform a full clean build without tests and benchmark:
make clean && make release
- Perform a full clean build with tests and benchmark (optional):
make clean && make all
- Run tests (optional):
make test && make pytest
For development, use make debug to build a non-optimized debug version.
To build in parallel, pass NUM_THREADS as parameter, e.g., make NUM_THREADS=8.
After build, our CLI binary kuzu_shell is available under the directory build/release/tools/shell/.
Precompiled binary of our latest release can be downloaded here.
Our Python package can be directly install through pip.
pip install kuzu
For more installation and usage instructions, please refer to our website.
We take tinysnb as an example graph, which is under dataset/demo-db/csv in our source code, and can be downloaded here.
./build/release/tools/shell/kuzu_shell -i "./testdb"
kuzu> CREATE NODE TABLE User(name STRING, age INT64, PRIMARY KEY (name));
kuzu> CREATE REL TABLE Follows(FROM User TO User, since INT64);kuzu> COPY User FROM "dataset/demo-db/csv/user.csv";
kuzu> COPY Follows FROM "dataset/demo-db/csv/follows.csv";After creating node/rel tables, and importing csv files, you can now run queries!
kuzu> MATCH (a:User)-[f:Follows]->(b:User) RETURN a.name, b.name, f.since;import kuzu
db = kuzu.Database('./testdb')
conn = kuzu.Connection(db)
conn.execute("CREATE NODE TABLE User(name STRING, age INT64, PRIMARY KEY (name))")
conn.execute("CREATE REL TABLE Follows(FROM User TO User, since INT64)")
conn.execute('COPY User FROM "user.csv"')
conn.execute('COPY Follows FROM "follows.csv"')
# Run a query.
results = conn.execute('MATCH (u:User) RETURN COUNT(*);')
while results.hasNext():
print(results.getNext())
# Run a query and get results as a pandas dataframe.
results = conn.execute('MATCH (a:User)-[f:Follows]->(b:User) RETURN a.name, f.since, b.name;').getAsDF()
print(results)
# Run a query and get results as an arrow table.
results = conn.execute('MATCH (u:User) RETURN u.name, u.age;').get_as_arrow(chunk_size=100)
print(results)Refer to our Data Import and Cypher section for more information.
We welcome contributions to Kùzu. If you are interested in contributing to Kùzu, please read our Contributing Guide.
By contributing to Kùzu, you agree that your contributions will be licensed under the MIT License.
If you are a researcher and use Kùzu in your work, we encourage you to cite our work. You can use the following BibTeX citation:
@inproceedings{kuzu:cidr,
author = {Xiyang Feng and
Guodong Jin and
Ziyi Chen and
Chang Liu and
Semih Saliho\u{g}lu},
title={K\`uzu Graph Database Management System},
booktitle={CIDR},
year={2023}
}
@misc{kuzu-github,
author = {Xiyang Feng and
Guodong Jin and
Ziyi Chen and
Chang Liu and
Semih Saliho\u{g}lu},
title = {{K\`uzu Database Management System Source Code}},
howpublished = {\url{https://github.com/kuzudb/kuzu}},
month = nov,
year = 2022
}
You can contact us at [email protected] or join our Slack workspace.