Guide for New Contributors¶
This is a guide for new contributors to the Lance project. Even if you have no previous experience with python, rust, and open source, you can still make an non-trivial impact by helping us improve documentation, examples, and more. For experienced developers, the issues you can work on run the gamut from warm-ups to serious challenges in python and rust.
If you have any questions, please join our Discord for real-time support. Your feedback is always welcome!
Join our Discord and say hi
Setup your development environment
Pick an issue to work on. See https://github.com/lancedb/lance/contribute for good first issues.
Currently Lance is implemented in Rust and comes with a Python wrapper. So you’ll want to make sure you setup both.
Fork the repo
Pick Github issue
Create a branch for the issue
Make your changes
Create a pull request from your fork to lancedb/lance
Get feedback and iterate
Go back to step 2
We use black isort and ruff for python and cargo fmt and clippy for rust.
To run the python linters:
ruff check python
To run the rust linters:
cargo fmt --all
cargo clippy --all-features --tests
The core format is implemented in Rust under the rust directory. Once you’ve setup Rust you can build the core format with:
This builds the debug build. For the optimized release build:
cargo build -r
To run the Rust unit tests:
If you’re working on a performance related feature, benchmarks can be run via:
The Python bindings for Lance uses a mix of Rust (pyo3) and Python. The Rust code that directly supports the Python bindings are under python/src while the pure Python code lives under python/python.
To build the Python bindings, first install requirements:
pip install maturin
To make a dev install:
To run the Python unit tests:
The documentation is built using Sphinx and lives under docs. To build the docs, first install requirements:
pip install -r docs/requirements.txt
Then build the docs:
The docs will be built under docs/build/html.
Example notebooks are under examples. These are standalone notebooks you should be able to download and run.
In python, Lance integrates with DuckDB via Apache Arrow. Outside of python, the highly experimental duckdb extension for Lance lives under integration/duckdb_lance. This uses the DuckDB Rust extension framework. The main code lives under integration/duckdb_lance/src. Follow the integration README for more details.
Our Rust benchmarks are run multiple times a day and the history can be found here.
Separately, we have vector index benchmarks that test against the sift1m dataset, as well as benchmarks for tpch. These live under benchmarks.
Code of Conduct¶
See https://www.python.org/psf/conduct/ and https://www.rust-lang.org/policies/code-of-conduct for details.