numru

crates.io

A high-performance numeric computation library written in Rust.

Motivation

Numru is a numeric computation library that aims to provide a high-performance, easy-to-use, and flexible API for numerical operations. It is inspired by NumPy, a popular numerical computation library in Python. Numru is designed to be a fundamental library for numeric computing with Rust.

Get Started

This getting started guide might change and should not be a source of absolute truth. Check the unit tests and in examples if you want to stay up to date with how things should be done. Some APIs will most likely be changed in the future.

[dependencies]
numru = "0.2.0"

And a simple code:

use numru::arr;
use std::f64::consts::{E, PI, TAU};

fn main() {
    let a = arr![42, -17, 256, 3, 99, -8];
    println!("a.shape() = {:?}", a.shape());
    a.visualize().execute();

    let b = arr![[TAU, -PI, 1.61], [E, 0.98, -7.42], [4.67, -0.45, 8.88]];
    println!("\nb.shape() = {:?}", b.shape());
    b.visualize()
        .decimal_points(1)
        .execute();

    let c = arr![
        [[101, 202, 303], [404, 505, 606]],
        [[-707, -808, -909], [111, 222, 333]]
    ];
    println!("\nc.shape() = {:?}", c.shape());
    c.visualize().execute();
}

Output of the code above:

a.shape() = Ix { dims: [6] }
[42, -17, 256, 3, 99, -8]

b.shape() = Ix { dims: [3, 3] }
[
   [6.3, -3.1, 1.6 ]
   [2.7, 1.0 , -7.4]
   [4.7, -0.5, 8.9 ]
]

c.shape() = Ix { dims: [2, 2, 3] }
[
   [
      [101 , 202 , 303 ]
      [404 , 505 , 606 ]
   ]
   [
      [-707, -808, -909]
      [111 , 222 , 333 ]
   ]
]

Features

Numru will offer a variety of different numerical operations and data types. It is intended to be a fundamental library for numeric computing with Rust.

Supported Data Types

Planned Data Types (Future)

i8, i16, i32, i128
u8, u16, u32, u64, u128
f32
bool
String, &str

Supported Operations

Note that currently we only show the numru equivalents as the ones that are planned. They do not exist yet.

Operation	Type	NumPy Equivalent	Numru Equivalent
Create Array	Array Creation	`np.array([1, 2, 3])`	`arr![1, 2, 3]`
Zeros Array	Array Creation	`np.zeros((3,3))`	`zeros!(i64, 3, 3)`
Ones Array	Array Creation	`np.ones((3,3))`	`ones!(i64, 3, 3)`
Arange	Array Creation	`np.arange(start, stop, step)`	🚧
Linspace	Array Creation	`np.linspace(start, stop, num)`	🚧
Mean	Reduction	`np.mean(a)`	`a.mean().compute()`
Min	Reduction	`np.min(a)`	`a.min().compute()`
Max	Reduction	`np.max(a)`	`a.max().compute()`
Dot Product	Linear Algebra	`np.dot(a, b)`	🚧
Reshape	Manipulation	`a.reshape((4, 3, 3))`	🚧
Concatenate	Manipulation	`np.concatenate([a, b], axis=0)`	🚧
Element-wise Add	Element-wise Ops	`a + b`	🚧
Element-wise Sub	Element-wise Ops	`a - b`	🚧
Element-wise Mul	Element-wise Ops	`a * b`	🚧
Element-wise Div	Element-wise Ops	`a / b`	🚧

Utility Features

These utility features help with visualization, debugging, array exploration and more.

Feature	Type	Numru	Description
Visualization	Visualization	`a.visualize().execute()`	Print an array in a human-readable format
Shape Inspection	Introspection	`a.shape()`	Get the shape of the array
Data Type Check	Introspection	`a.dtype()`	Retrieve the data type of the array

License

The MIT License.

numru 0.2.0