glem

Using matrices from the glam crate, perform linear algebra transforms by chaining adaptors. Derive the inverse function for a chain of adaptors, without having to compute the matrix inverse.

Instead of passing around rotation or translation transformation matrices, just pass around the information to build these matrices. Then only when glem::build() is called actually build the matrix. This likely avoids a lot of temporary matrices being created and destroyed when executing long sequences of transformations.

You can find glem on github and crates.io. Documentation at docs.rs

Example

#[test]
fn example() {
    use glem::prelude::*;

    let c = glem::rotate_x(0.5).chain(glem::translate(55.0, -5.0, -6.0));

    let c = glem::build(c);

    let x = glem::build(glem::rotate_x(0.5));
    let y = glem::build(glem::translate(55.0, -5.0, -6.0));

    assert_eq!(c, x * y);
}

Inverse Example

#[test]
fn inverse_example() {
    let c = glem::combine!(
        glem::rotate_x(0.5),
        glem::rotate_y(0.2),
        glem::rotate_z(0.1),
        glem::translate(55.0, -5.0, -6.0),
        glem::scale(2.0, 4.0, -2.0)
    );

    let c2 = glem::combine!(
        glem::scale(1.0 / 2.0, 1.0 / 4.0, -1.0 / 2.0),
        glem::translate(-55.0, 5.0, 6.0),
        glem::rotate_z(-0.1),
        glem::rotate_y(-0.2),
        glem::rotate_x(-0.5)
    );

    assert_eq!(glem::build_inverse(&c), glem::build(c2));

    approx_eq(&glem::build(&c).inverse(), &glem::build_inverse(c));
}

fn approx_eq(a: &glam::f32::Mat4, b: &glam::f32::Mat4) {
    let a = a.to_cols_array();
    let b = b.to_cols_array();

    for (a, b) in a.into_iter().zip(b.into_iter()) {
        assert!((a - b).abs() < 0.000001, "{}:{}", a, b);
    }
}