rust-numpy

Rust bindings for the NumPy C-API.

API documentation

Requirements

Rust >= 1.48.0
- Basically, our MSRV follows the one of PyO3
Python >= 3.7
- Python 3.6 support was dropped from 0.16
Some Rust libraries
- ndarray for Rust-side matrix library
- PyO3 for Python bindings
- And more (see Cargo.toml)
numpy installed in your Python environments (e.g., via pip install numpy)
- We recommend numpy >= 1.16.0, though older versions may work

Example

Write a Python module in Rust

Please see the simple example for how to get started.

There are also examples using ndarray-linalg and rayon.

[lib]
name = "rust_ext"
crate-type = ["cdylib"]

[dependencies]
pyo3 = { version = "0.18", features = ["extension-module"] }
numpy = "0.18"

use numpy::ndarray::{ArrayD, ArrayViewD, ArrayViewMutD};
use numpy::{IntoPyArray, PyArrayDyn, PyReadonlyArrayDyn};
use pyo3::{pymodule, types::PyModule, PyResult, Python};

#[pymodule]
fn rust_ext(_py: Python<'_>, m: &PyModule) -> PyResult<()> {
    // example using immutable borrows producing a new array
    fn axpy(a: f64, x: ArrayViewD<'_, f64>, y: ArrayViewD<'_, f64>) -> ArrayD<f64> {
        a * &x + &y
    }

    // example using a mutable borrow to modify an array in-place
    fn mult(a: f64, mut x: ArrayViewMutD<'_, f64>) {
        x *= a;
    }

    // wrapper of `axpy`
    #[pyfn(m)]
    #[pyo3(name = "axpy")]
    fn axpy_py<'py>(
        py: Python<'py>,
        a: f64,
        x: PyReadonlyArrayDyn<f64>,
        y: PyReadonlyArrayDyn<f64>,
    ) -> &'py PyArrayDyn<f64> {
        let x = x.as_array();
        let y = y.as_array();
        let z = axpy(a, x, y);
        z.into_pyarray(py)
    }

    // wrapper of `mult`
    #[pyfn(m)]
    #[pyo3(name = "mult")]
    fn mult_py(_py: Python<'_>, a: f64, x: &PyArrayDyn<f64>) {
        let x = unsafe { x.as_array_mut() };
        mult(a, x);
    }

    Ok(())
}

Execute a Python program from Rust and get results

[package]
name = "numpy-test"

[dependencies]
pyo3 = { version = "0.18", features = ["auto-initialize"] }
numpy = "0.18"

use numpy::PyArray1;
use pyo3::{types::IntoPyDict, PyResult, Python};

fn main() -> PyResult<()> {
    Python::with_gil(|py| {
        let np = py.import("numpy")?;
        let locals = [("np", np)].into_py_dict(py);

        let pyarray: &PyArray1<i32> = py
            .eval("np.absolute(np.array([-1, -2, -3], dtype='int32'))", Some(locals), None)?
            .extract()?;

        let readonly = pyarray.readonly();
        let slice = readonly.as_slice()?;
        assert_eq!(slice, &[1, 2, 3]);

        Ok(())
    })
}

Dependency on ndarray

This crate uses types from ndarray in its public API. ndarray is re-exported in the crate root so that you do not need to specify it as a direct dependency.

Furthermore, this crate is compatible with multiple versions of ndarray and therefore depends on a range of semver-incompatible versions, currently >= 0.13, < 0.16. Cargo does not automatically choose a single version of ndarray by itself if you depend directly or indirectly on anything but that exact range. It can therefore be necessary to manually unify these dependencies.

For example, if you specify the following dependencies

numpy = "0.18"
ndarray = "0.13"

this will currently depend on both version 0.13.1 and 0.15.3 of ndarray by default even though 0.13.1 is within the range >= 0.13, < 0.16. To fix this, you can run

cargo update --package ndarray:0.15.3 --precise 0.13.1

to achieve a single dependency on version 0.13.1 of ndarray.

Contributing

We welcome issues and pull requests.

PyO3's Contributing.md is a nice guide for starting.

Also, we have a Gitter channel for communicating.

PyO3/rust-numpy

PyO3

Reviews

Repository Details