#mat #serialization #iterator #row #b


A lightweight, rusty matrix library that allows for simple and fast matrix operations

2 unstable releases

0.2.0 May 28, 2023
0.1.0 May 25, 2023

#551 in Algorithms

MIT license

474 lines


A lightweight, rusty matrix library that allows for simple (as in, easy to put into place) matrix handling and operations.
Operations are not done in place, so most functions, especially mathematical operations, are pure functions that consume their arguments and output a new matrix.

build-badge GitHub GitHub release (latest by date) Twitter URL


  • Matrix initialization (filled, random, custom)
  • Basic operations Mat with Mat and Mat with Scalar
  • Better error handling (Results for operations that may fail because of uncompatible sizes?)
  • Matrix rotations
  • Macros for simpler initialization
  • Better recursive matrices (operations, display, etc.)
  • Computations on GPU?

Notes on usage

Do not hesitate to go see tests.rs for more examples.

Creating matrices

Declaring a matrix is rather straightforward: you only need to specify the type of elements, the number of rows and the number of columns.

let mut a = Matrix::<f64>::new(2, 3);

By default, using Matrix::new() will initialize the matrix with zeros (it is thus only available with T: num::Num, so a number). If needed, you can specify the whole content of the matrix using Matrix::from<Vec<Vec<T>>>(), so with a vector of vectors (with each representing a row).

// 1 2 3
// 4 5 6
let a = Matrix::<f64>::from(vec![
        vec![1.0f64, 2.0f64, 3.0f64],
        vec![4.0f64, 5.0f64, 6.0f64]


Multiplication and addition are implemented between matrices, and between a matrix and an object of type T (the type of the elements) if T is a number.
Matrix-matrix operations may require compatibility between the two (sizewise) ; operations on matrices, like the dot product, return a Result<> since the size checks mostly happen at runtime for now.
As such, inline arithmetics is discouraged: a + b * (d - e) should rather be computed step by step as good pratice, or else all 'em unwraps will make things unreadable.

No operation is done in-place: they all generate a new matrix. You'll need to explicitly .clone() a matrix if it should be used in several operations.

Current implemented operations are as follows:

  • Mat * Mat and Mat * scal
  • Mat / Mat and Mat / scal
  • Mat + Mat and Mat + scal
  • Mat - Mat and Mat - scal

Summing up a matrix's content is also available for all types that implement T+T and can be summed through an iterator: .sum()


You can apply a closure on each matrix elements using the .map() method


~49K SLoC