1050 lines
29 KiB
Rust
1050 lines
29 KiB
Rust
use num::{One, Zero};
|
||
use std::cmp::Ordering;
|
||
|
||
use na::dimension::{U15, U8};
|
||
use na::{
|
||
self, DMatrix, DVector, Matrix2, Matrix2x3, Matrix2x4, Matrix3, Matrix3x2, Matrix3x4, Matrix4,
|
||
Matrix4x3, Matrix4x5, Matrix5, Matrix6, MatrixMN, Real, RowVector3, RowVector4, RowVector5,
|
||
Vector1, Vector2, Vector3, Vector4, Vector5, Vector6,
|
||
};
|
||
|
||
#[test]
|
||
fn iter() {
|
||
let a = Matrix2x3::new(1.0, 2.0, 3.0, 4.0, 5.0, 6.0);
|
||
|
||
let mut it = a.iter();
|
||
assert_eq!(*it.next().unwrap(), 1.0);
|
||
assert_eq!(*it.next().unwrap(), 4.0);
|
||
assert_eq!(*it.next().unwrap(), 2.0);
|
||
assert_eq!(*it.next().unwrap(), 5.0);
|
||
assert_eq!(*it.next().unwrap(), 3.0);
|
||
assert_eq!(*it.next().unwrap(), 6.0);
|
||
assert!(it.next().is_none());
|
||
|
||
let row = a.row(0);
|
||
let mut it = row.iter();
|
||
assert_eq!(*it.next().unwrap(), 1.0);
|
||
assert_eq!(*it.next().unwrap(), 2.0);
|
||
assert_eq!(*it.next().unwrap(), 3.0);
|
||
assert!(it.next().is_none());
|
||
|
||
let row = a.row(1);
|
||
let mut it = row.iter();
|
||
assert_eq!(*it.next().unwrap(), 4.0);
|
||
assert_eq!(*it.next().unwrap(), 5.0);
|
||
assert_eq!(*it.next().unwrap(), 6.0);
|
||
assert!(it.next().is_none());
|
||
|
||
let m22 = row.column(1);
|
||
let mut it = m22.iter();
|
||
assert_eq!(*it.next().unwrap(), 5.0);
|
||
assert!(it.next().is_none());
|
||
|
||
let col = a.column(0);
|
||
let mut it = col.iter();
|
||
assert_eq!(*it.next().unwrap(), 1.0);
|
||
assert_eq!(*it.next().unwrap(), 4.0);
|
||
assert!(it.next().is_none());
|
||
|
||
let col = a.column(1);
|
||
let mut it = col.iter();
|
||
assert_eq!(*it.next().unwrap(), 2.0);
|
||
assert_eq!(*it.next().unwrap(), 5.0);
|
||
assert!(it.next().is_none());
|
||
|
||
let col = a.column(2);
|
||
let mut it = col.iter();
|
||
assert_eq!(*it.next().unwrap(), 3.0);
|
||
assert_eq!(*it.next().unwrap(), 6.0);
|
||
assert!(it.next().is_none());
|
||
}
|
||
|
||
#[test]
|
||
fn debug_output_corresponds_to_data_container() {
|
||
assert!(
|
||
format!("{:?}", Matrix2::new(1.0, 2.0, 3.0, 4.0)) == "Matrix { data: [1, 3, 2, 4] }" || // Current output on the stable chanel.
|
||
format!("{:?}", Matrix2::new(1.0, 2.0, 3.0, 4.0)) == "Matrix { data: [1.0, 3.0, 2.0, 4.0] }" // Current output on the nightyl chanel.
|
||
);
|
||
}
|
||
|
||
#[test]
|
||
fn is_column_major() {
|
||
let a = Matrix2x3::new(1.0, 2.0, 3.0, 4.0, 5.0, 6.0);
|
||
|
||
let expected = &[1.0, 4.0, 2.0, 5.0, 3.0, 6.0];
|
||
|
||
assert_eq!(a.as_slice(), expected);
|
||
|
||
let a = Matrix2x3::from_row_slice(&[1.0, 2.0, 3.0, 4.0, 5.0, 6.0]);
|
||
|
||
assert_eq!(a.as_slice(), expected);
|
||
|
||
let a = Matrix2x3::from_column_slice(&[1.0, 4.0, 2.0, 5.0, 3.0, 6.0]);
|
||
|
||
assert_eq!(a.as_slice(), expected);
|
||
}
|
||
|
||
#[test]
|
||
fn linear_index() {
|
||
let a = Matrix2x3::new(1, 2, 3, 4, 5, 6);
|
||
|
||
assert_eq!(a[0], 1);
|
||
assert_eq!(a[1], 4);
|
||
assert_eq!(a[2], 2);
|
||
assert_eq!(a[3], 5);
|
||
assert_eq!(a[4], 3);
|
||
assert_eq!(a[5], 6);
|
||
|
||
let b = Vector4::new(1, 2, 3, 4);
|
||
|
||
assert_eq!(b[0], 1);
|
||
assert_eq!(b[1], 2);
|
||
assert_eq!(b[2], 3);
|
||
assert_eq!(b[3], 4);
|
||
|
||
let c = RowVector4::new(1, 2, 3, 4);
|
||
|
||
assert_eq!(c[0], 1);
|
||
assert_eq!(c[1], 2);
|
||
assert_eq!(c[2], 3);
|
||
assert_eq!(c[3], 4);
|
||
}
|
||
|
||
#[test]
|
||
fn identity() {
|
||
let id1 = Matrix3::<f64>::identity();
|
||
let id2 = Matrix3x4::new(1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0);
|
||
let id2bis = Matrix3x4::identity();
|
||
let id3 = Matrix4x3::new(1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0);
|
||
let id3bis = Matrix4x3::identity();
|
||
|
||
let not_id1 = Matrix3::identity() * 2.0;
|
||
let not_id2 = Matrix3x4::new(1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 0.0, 1.0, 1.0);
|
||
let not_id3 = Matrix4x3::new(1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 0.0, 0.0, 1.0, 0.0, 1.0, 0.0);
|
||
|
||
assert_eq!(id2, id2bis);
|
||
assert_eq!(id3, id3bis);
|
||
assert!(id1.is_identity(0.0));
|
||
assert!(id2.is_identity(0.0));
|
||
assert!(id3.is_identity(0.0));
|
||
assert!(!not_id1.is_identity(0.0));
|
||
assert!(!not_id2.is_identity(0.0));
|
||
assert!(!not_id3.is_identity(0.0));
|
||
}
|
||
|
||
#[test]
|
||
fn coordinates() {
|
||
let a = Matrix3x4::new(11, 12, 13, 14, 21, 22, 23, 24, 31, 32, 33, 34);
|
||
|
||
assert_eq!(a.m11, 11);
|
||
assert_eq!(a.m12, 12);
|
||
assert_eq!(a.m13, 13);
|
||
assert_eq!(a.m14, 14);
|
||
|
||
assert_eq!(a.m21, 21);
|
||
assert_eq!(a.m22, 22);
|
||
assert_eq!(a.m23, 23);
|
||
assert_eq!(a.m24, 24);
|
||
|
||
assert_eq!(a.m31, 31);
|
||
assert_eq!(a.m32, 32);
|
||
assert_eq!(a.m33, 33);
|
||
assert_eq!(a.m34, 34);
|
||
}
|
||
|
||
#[test]
|
||
fn from_diagonal() {
|
||
let diag = Vector3::new(1, 2, 3);
|
||
let expected = Matrix3::new(1, 0, 0, 0, 2, 0, 0, 0, 3);
|
||
let a = Matrix3::from_diagonal(&diag);
|
||
|
||
assert_eq!(a, expected);
|
||
}
|
||
|
||
#[test]
|
||
fn from_rows() {
|
||
let rows = &[
|
||
RowVector4::new(11, 12, 13, 14),
|
||
RowVector4::new(21, 22, 23, 24),
|
||
RowVector4::new(31, 32, 33, 34),
|
||
];
|
||
|
||
let expected = Matrix3x4::new(11, 12, 13, 14, 21, 22, 23, 24, 31, 32, 33, 34);
|
||
|
||
let a = Matrix3x4::from_rows(rows);
|
||
|
||
assert_eq!(a, expected);
|
||
}
|
||
|
||
#[test]
|
||
fn from_columns() {
|
||
let columns = &[
|
||
Vector3::new(11, 21, 31),
|
||
Vector3::new(12, 22, 32),
|
||
Vector3::new(13, 23, 33),
|
||
Vector3::new(14, 24, 34),
|
||
];
|
||
|
||
let expected = Matrix3x4::new(11, 12, 13, 14, 21, 22, 23, 24, 31, 32, 33, 34);
|
||
|
||
let a = Matrix3x4::from_columns(columns);
|
||
|
||
assert_eq!(a, expected);
|
||
}
|
||
|
||
#[test]
|
||
fn from_columns_dynamic() {
|
||
let columns = &[
|
||
DVector::from_row_slice(3, &[11, 21, 31]),
|
||
DVector::from_row_slice(3, &[12, 22, 32]),
|
||
DVector::from_row_slice(3, &[13, 23, 33]),
|
||
DVector::from_row_slice(3, &[14, 24, 34]),
|
||
];
|
||
|
||
let expected = DMatrix::from_row_slice(3, 4, &[11, 12, 13, 14, 21, 22, 23, 24, 31, 32, 33, 34]);
|
||
|
||
let a = DMatrix::from_columns(columns);
|
||
|
||
assert_eq!(a, expected);
|
||
}
|
||
|
||
#[test]
|
||
#[should_panic]
|
||
fn from_too_many_rows() {
|
||
let rows = &[
|
||
RowVector4::new(11, 12, 13, 14),
|
||
RowVector4::new(21, 22, 23, 24),
|
||
RowVector4::new(31, 32, 33, 34),
|
||
RowVector4::new(31, 32, 33, 34),
|
||
];
|
||
|
||
let _ = Matrix3x4::from_rows(rows);
|
||
}
|
||
|
||
#[test]
|
||
#[should_panic]
|
||
fn from_not_enough_columns() {
|
||
let columns = &[Vector3::new(11, 21, 31), Vector3::new(14, 24, 34)];
|
||
|
||
let _ = Matrix3x4::from_columns(columns);
|
||
}
|
||
|
||
#[test]
|
||
#[should_panic]
|
||
fn from_rows_with_different_dimensions() {
|
||
let columns = &[
|
||
DVector::from_row_slice(3, &[11, 21, 31]),
|
||
DVector::from_row_slice(3, &[12, 22, 32, 33]),
|
||
];
|
||
|
||
let _ = DMatrix::from_columns(columns);
|
||
}
|
||
|
||
#[test]
|
||
fn copy_from_slice() {
|
||
let mut a = Matrix3::zeros();
|
||
let data = [1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0];
|
||
let expected_a = Matrix3::new(1.0, 4.0, 7.0, 2.0, 5.0, 8.0, 3.0, 6.0, 9.0);
|
||
|
||
a.copy_from_slice(&data);
|
||
|
||
assert_eq!(a, expected_a);
|
||
}
|
||
|
||
#[should_panic]
|
||
#[test]
|
||
fn copy_from_slice_too_small() {
|
||
let mut a = Matrix3::zeros();
|
||
let data = [1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0];
|
||
a.copy_from_slice(&data);
|
||
}
|
||
|
||
#[should_panic]
|
||
#[test]
|
||
fn copy_from_slice_too_large() {
|
||
let mut a = Matrix3::zeros();
|
||
let data = [1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0, 10.0];
|
||
a.copy_from_slice(&data);
|
||
}
|
||
|
||
#[test]
|
||
fn to_homogeneous() {
|
||
let a = Vector3::new(1.0, 2.0, 3.0);
|
||
let expected_a = Vector4::new(1.0, 2.0, 3.0, 0.0);
|
||
|
||
let b = DVector::from_row_slice(3, &[1.0, 2.0, 3.0]);
|
||
let expected_b = DVector::from_row_slice(4, &[1.0, 2.0, 3.0, 0.0]);
|
||
|
||
let c = Matrix2::new(1.0, 2.0, 3.0, 4.0);
|
||
let expected_c = Matrix3::new(1.0, 2.0, 0.0, 3.0, 4.0, 0.0, 0.0, 0.0, 1.0);
|
||
|
||
let d = DMatrix::from_row_slice(2, 2, &[1.0, 2.0, 3.0, 4.0]);
|
||
let expected_d = DMatrix::from_row_slice(3, 3, &[1.0, 2.0, 0.0, 3.0, 4.0, 0.0, 0.0, 0.0, 1.0]);
|
||
|
||
assert_eq!(a.to_homogeneous(), expected_a);
|
||
assert_eq!(b.to_homogeneous(), expected_b);
|
||
assert_eq!(c.to_homogeneous(), expected_c);
|
||
assert_eq!(d.to_homogeneous(), expected_d);
|
||
}
|
||
|
||
#[test]
|
||
fn push() {
|
||
let a = Vector3::new(1.0, 2.0, 3.0);
|
||
let expected_a = Vector4::new(1.0, 2.0, 3.0, 4.0);
|
||
|
||
let b = DVector::from_row_slice(3, &[1.0, 2.0, 3.0]);
|
||
let expected_b = DVector::from_row_slice(4, &[1.0, 2.0, 3.0, 4.0]);
|
||
|
||
assert_eq!(a.push(4.0), expected_a);
|
||
assert_eq!(b.push(4.0), expected_b);
|
||
}
|
||
|
||
#[test]
|
||
fn simple_add() {
|
||
let a = Matrix2x3::new(1.0, 2.0, 3.0, 4.0, 5.0, 6.0);
|
||
|
||
let b = Matrix2x3::new(10.0, 20.0, 30.0, 40.0, 50.0, 60.0);
|
||
let c = DMatrix::from_row_slice(2, 3, &[10.0, 20.0, 30.0, 40.0, 50.0, 60.0]);
|
||
|
||
let expected = Matrix2x3::new(11.0, 22.0, 33.0, 44.0, 55.0, 66.0);
|
||
|
||
assert_eq!(expected, &a + &b);
|
||
assert_eq!(expected, &a + b);
|
||
assert_eq!(expected, a + &b);
|
||
assert_eq!(expected, a + b);
|
||
|
||
// Sum of a static matrix with a dynamic one.
|
||
assert_eq!(expected, &a + &c);
|
||
assert_eq!(expected, a + &c);
|
||
assert_eq!(expected, &c + &a);
|
||
assert_eq!(expected, &c + a);
|
||
}
|
||
|
||
#[test]
|
||
fn simple_sum() {
|
||
type M = Matrix2x3<f32>;
|
||
|
||
let a = M::new(1.0, 2.0, 3.0, 4.0, 5.0, 6.0);
|
||
let b = M::new(10.0, 20.0, 30.0, 40.0, 50.0, 60.0);
|
||
let c = M::new(100.0, 200.0, 300.0, 400.0, 500.0, 600.0);
|
||
|
||
assert_eq!(M::zero(), Vec::<M>::new().iter().sum());
|
||
assert_eq!(M::zero(), Vec::<M>::new().into_iter().sum());
|
||
assert_eq!(a + b, vec![a, b].iter().sum());
|
||
assert_eq!(a + b, vec![a, b].into_iter().sum());
|
||
assert_eq!(a + b + c, vec![a, b, c].iter().sum());
|
||
assert_eq!(a + b + c, vec![a, b, c].into_iter().sum());
|
||
}
|
||
|
||
#[test]
|
||
fn simple_scalar_mul() {
|
||
let a = Matrix2x3::new(1.0, 2.0, 3.0, 4.0, 5.0, 6.0);
|
||
|
||
let expected = Matrix2x3::new(10.0, 20.0, 30.0, 40.0, 50.0, 60.0);
|
||
|
||
assert_eq!(expected, a * 10.0);
|
||
assert_eq!(expected, &a * 10.0);
|
||
assert_eq!(expected, 10.0 * a);
|
||
assert_eq!(expected, 10.0 * &a);
|
||
}
|
||
|
||
#[test]
|
||
fn simple_mul() {
|
||
let a = Matrix2x3::new(1.0, 2.0, 3.0, 4.0, 5.0, 6.0);
|
||
|
||
let b = Matrix3x4::new(
|
||
10.0, 20.0, 30.0, 40.0, 50.0, 60.0, 70.0, 80.0, 90.0, 100.0, 110.0, 120.0,
|
||
);
|
||
|
||
let expected = Matrix2x4::new(380.0, 440.0, 500.0, 560.0, 830.0, 980.0, 1130.0, 1280.0);
|
||
|
||
assert_eq!(expected, &a * &b);
|
||
assert_eq!(expected, a * &b);
|
||
assert_eq!(expected, &a * b);
|
||
assert_eq!(expected, a * b);
|
||
}
|
||
|
||
#[test]
|
||
fn simple_product() {
|
||
type M = Matrix3<f32>;
|
||
|
||
let a = M::new(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0);
|
||
let b = M::new(10.0, 20.0, 30.0, 40.0, 50.0, 60.0, 70.0, 80.0, 90.0);
|
||
let c = M::new(
|
||
100.0, 200.0, 300.0, 400.0, 500.0, 600.0, 700.0, 800.0, 900.0,
|
||
);
|
||
|
||
assert_eq!(M::one(), Vec::<M>::new().iter().product());
|
||
assert_eq!(M::one(), Vec::<M>::new().into_iter().product());
|
||
assert_eq!(a * b, vec![a, b].iter().product());
|
||
assert_eq!(a * b, vec![a, b].into_iter().product());
|
||
assert_eq!(a * b * c, vec![a, b, c].iter().product());
|
||
assert_eq!(a * b * c, vec![a, b, c].into_iter().product());
|
||
}
|
||
|
||
#[test]
|
||
fn cross_product_vector_and_row_vector() {
|
||
let v1 = Vector3::new(1.0, 2.0, 3.0);
|
||
let v2 = Vector3::new(1.0, 5.0, 7.0);
|
||
let column_cross = v1.cross(&v2);
|
||
assert_eq!(column_cross, Vector3::new(-1.0, -4.0, 3.0));
|
||
|
||
let v1 = RowVector3::new(1.0, 2.0, 3.0);
|
||
let v2 = RowVector3::new(1.0, 5.0, 7.0);
|
||
let row_cross = v1.cross(&v2);
|
||
assert_eq!(row_cross, RowVector3::new(-1.0, -4.0, 3.0));
|
||
|
||
assert_eq!(
|
||
Vector3::new(1.0, 1.0, 0.0)
|
||
.cross(&Vector3::new(-0.5, 17.0, 0.0))
|
||
.transpose(),
|
||
RowVector3::new(1.0, 1.0, 0.0).cross(&RowVector3::new(-0.5, 17.0, 0.0))
|
||
);
|
||
}
|
||
|
||
#[test]
|
||
fn simple_scalar_conversion() {
|
||
let a = Matrix2x3::new(1.0, 2.0, 3.0, 4.0, 5.0, 6.0);
|
||
let expected = Matrix2x3::new(1, 2, 3, 4, 5, 6);
|
||
|
||
let a_u32: Matrix2x3<u32> = na::try_convert(a).unwrap(); // f32 -> u32
|
||
let a_f32: Matrix2x3<f32> = na::convert(a_u32); // u32 -> f32
|
||
|
||
assert_eq!(a, a_f32);
|
||
assert_eq!(expected, a_u32);
|
||
}
|
||
|
||
#[test]
|
||
fn apply() {
|
||
let mut a = Matrix4::new(
|
||
1.1, 2.2, 3.3, 4.4, 5.5, 6.6, 7.7, 8.8, 9.9, 8.8, 7.7, 6.6, 5.5, 4.4, 3.3, 2.2,
|
||
);
|
||
|
||
let expected = Matrix4::new(
|
||
1.0, 2.0, 3.0, 4.0, 6.0, 7.0, 8.0, 9.0, 10.0, 9.0, 8.0, 7.0, 6.0, 4.0, 3.0, 2.0,
|
||
);
|
||
|
||
a.apply(|e| e.round());
|
||
|
||
assert_eq!(a, expected);
|
||
}
|
||
|
||
#[test]
|
||
fn map() {
|
||
let a = Matrix4::new(
|
||
1.1f64, 2.2, 3.3, 4.4, 5.5, 6.6, 7.7, 8.8, 9.9, 8.8, 7.7, 6.6, 5.5, 4.4, 3.3, 2.2,
|
||
);
|
||
|
||
let expected = Matrix4::new(1, 2, 3, 4, 6, 7, 8, 9, 10, 9, 8, 7, 6, 4, 3, 2);
|
||
|
||
let computed = a.map(|e| e.round() as i64);
|
||
|
||
assert_eq!(computed, expected);
|
||
}
|
||
|
||
#[test]
|
||
fn map_with_location() {
|
||
let a = Matrix4::new(1, 1, 1, 1, 2, 2, 2, 2, 3, 3, 3, 3, 4, 4, 4, 4);
|
||
|
||
let expected = Matrix4::new(1, 2, 3, 4, 3, 4, 5, 6, 5, 6, 7, 8, 7, 8, 9, 10);
|
||
|
||
let computed = a.map_with_location(|i, j, e| e + i + j);
|
||
|
||
assert_eq!(computed, expected);
|
||
}
|
||
|
||
#[test]
|
||
fn zip_map() {
|
||
let a = Matrix3::new(11i32, 12, 13, 21, 22, 23, 31, 32, 33);
|
||
|
||
let b = Matrix3::new(11u32, 12, 13, 21, 22, 23, 31, 32, 33);
|
||
|
||
let expected = Matrix3::new(22.0f32, 24.0, 26.0, 42.0, 44.0, 46.0, 62.0, 64.0, 66.0);
|
||
|
||
let computed = a.zip_map(&b, |ea, eb| ea as f32 + eb as f32);
|
||
|
||
assert_eq!(computed, expected);
|
||
}
|
||
|
||
#[test]
|
||
#[should_panic]
|
||
fn trace_panic() {
|
||
let m = DMatrix::<f32>::new_random(2, 3);
|
||
let _ = m.trace();
|
||
}
|
||
|
||
#[test]
|
||
fn trace() {
|
||
let m = Matrix2::new(1.0, 20.0, 30.0, 4.0);
|
||
assert_eq!(m.trace(), 5.0);
|
||
}
|
||
|
||
#[test]
|
||
fn simple_transpose() {
|
||
let a = Matrix2x3::new(1.0, 2.0, 3.0, 4.0, 5.0, 6.0);
|
||
let expected = Matrix3x2::new(1.0, 4.0, 2.0, 5.0, 3.0, 6.0);
|
||
|
||
assert_eq!(a.transpose(), expected);
|
||
}
|
||
|
||
#[test]
|
||
fn simple_transpose_mut() {
|
||
let mut a = Matrix3::new(1.0, 2.0, 3.0, 4.0, 5.0, 6.0, 7.0, 8.0, 9.0);
|
||
let expected = Matrix3::new(1.0, 4.0, 7.0, 2.0, 5.0, 8.0, 3.0, 6.0, 9.0);
|
||
|
||
a.transpose_mut();
|
||
assert_eq!(a, expected);
|
||
}
|
||
|
||
#[test]
|
||
fn vector_index_mut() {
|
||
let mut v = Vector3::new(1, 2, 3);
|
||
|
||
assert_eq!(v[0], 1);
|
||
assert_eq!(v[1], 2);
|
||
assert_eq!(v[2], 3);
|
||
|
||
v[0] = 10;
|
||
v[1] = 20;
|
||
v[2] = 30;
|
||
|
||
assert_eq!(v, Vector3::new(10, 20, 30));
|
||
}
|
||
|
||
#[test]
|
||
fn components_mut() {
|
||
let mut m2 = Matrix2::from_element(1.0);
|
||
let mut m3 = Matrix3::from_element(1.0);
|
||
let mut m4 = Matrix4::from_element(1.0);
|
||
let mut m5 = Matrix5::from_element(1.0);
|
||
let mut m6 = Matrix6::from_element(1.0);
|
||
|
||
m2.m11 = 0.0;
|
||
m2.m12 = 0.0;
|
||
m2.m21 = 0.0;
|
||
m2.m22 = 0.0;
|
||
|
||
m3.m11 = 0.0;
|
||
m3.m12 = 0.0;
|
||
m3.m13 = 0.0;
|
||
m3.m21 = 0.0;
|
||
m3.m22 = 0.0;
|
||
m3.m23 = 0.0;
|
||
m3.m31 = 0.0;
|
||
m3.m32 = 0.0;
|
||
m3.m33 = 0.0;
|
||
|
||
m4.m11 = 0.0;
|
||
m4.m12 = 0.0;
|
||
m4.m13 = 0.0;
|
||
m4.m14 = 0.0;
|
||
m4.m21 = 0.0;
|
||
m4.m22 = 0.0;
|
||
m4.m23 = 0.0;
|
||
m4.m24 = 0.0;
|
||
m4.m31 = 0.0;
|
||
m4.m32 = 0.0;
|
||
m4.m33 = 0.0;
|
||
m4.m34 = 0.0;
|
||
m4.m41 = 0.0;
|
||
m4.m42 = 0.0;
|
||
m4.m43 = 0.0;
|
||
m4.m44 = 0.0;
|
||
|
||
m5.m11 = 0.0;
|
||
m5.m12 = 0.0;
|
||
m5.m13 = 0.0;
|
||
m5.m14 = 0.0;
|
||
m5.m15 = 0.0;
|
||
m5.m21 = 0.0;
|
||
m5.m22 = 0.0;
|
||
m5.m23 = 0.0;
|
||
m5.m24 = 0.0;
|
||
m5.m25 = 0.0;
|
||
m5.m31 = 0.0;
|
||
m5.m32 = 0.0;
|
||
m5.m33 = 0.0;
|
||
m5.m34 = 0.0;
|
||
m5.m35 = 0.0;
|
||
m5.m41 = 0.0;
|
||
m5.m42 = 0.0;
|
||
m5.m43 = 0.0;
|
||
m5.m44 = 0.0;
|
||
m5.m45 = 0.0;
|
||
m5.m51 = 0.0;
|
||
m5.m52 = 0.0;
|
||
m5.m53 = 0.0;
|
||
m5.m54 = 0.0;
|
||
m5.m55 = 0.0;
|
||
|
||
m6.m11 = 0.0;
|
||
m6.m12 = 0.0;
|
||
m6.m13 = 0.0;
|
||
m6.m14 = 0.0;
|
||
m6.m15 = 0.0;
|
||
m6.m16 = 0.0;
|
||
m6.m21 = 0.0;
|
||
m6.m22 = 0.0;
|
||
m6.m23 = 0.0;
|
||
m6.m24 = 0.0;
|
||
m6.m25 = 0.0;
|
||
m6.m26 = 0.0;
|
||
m6.m31 = 0.0;
|
||
m6.m32 = 0.0;
|
||
m6.m33 = 0.0;
|
||
m6.m34 = 0.0;
|
||
m6.m35 = 0.0;
|
||
m6.m36 = 0.0;
|
||
m6.m41 = 0.0;
|
||
m6.m42 = 0.0;
|
||
m6.m43 = 0.0;
|
||
m6.m44 = 0.0;
|
||
m6.m45 = 0.0;
|
||
m6.m46 = 0.0;
|
||
m6.m51 = 0.0;
|
||
m6.m52 = 0.0;
|
||
m6.m53 = 0.0;
|
||
m6.m54 = 0.0;
|
||
m6.m55 = 0.0;
|
||
m6.m56 = 0.0;
|
||
m6.m61 = 0.0;
|
||
m6.m62 = 0.0;
|
||
m6.m63 = 0.0;
|
||
m6.m64 = 0.0;
|
||
m6.m65 = 0.0;
|
||
m6.m66 = 0.0;
|
||
|
||
assert!(m2.is_zero());
|
||
assert!(m3.is_zero());
|
||
assert!(m4.is_zero());
|
||
assert!(m5.is_zero());
|
||
assert!(m6.is_zero());
|
||
|
||
let mut v1 = Vector1::from_element(1.0);
|
||
let mut v2 = Vector2::from_element(1.0);
|
||
let mut v3 = Vector3::from_element(1.0);
|
||
let mut v4 = Vector4::from_element(1.0);
|
||
let mut v5 = Vector5::from_element(1.0);
|
||
let mut v6 = Vector6::from_element(1.0);
|
||
|
||
v1.x = 0.0;
|
||
v2.x = 0.0;
|
||
v2.y = 0.0;
|
||
v3.x = 0.0;
|
||
v3.y = 0.0;
|
||
v3.z = 0.0;
|
||
v4.x = 0.0;
|
||
v4.y = 0.0;
|
||
v4.z = 0.0;
|
||
v4.w = 0.0;
|
||
v5.x = 0.0;
|
||
v5.y = 0.0;
|
||
v5.z = 0.0;
|
||
v5.w = 0.0;
|
||
v5.a = 0.0;
|
||
v6.x = 0.0;
|
||
v6.y = 0.0;
|
||
v6.z = 0.0;
|
||
v6.w = 0.0;
|
||
v6.a = 0.0;
|
||
v6.b = 0.0;
|
||
|
||
assert!(v1.is_zero());
|
||
assert!(v2.is_zero());
|
||
assert!(v3.is_zero());
|
||
assert!(v4.is_zero());
|
||
assert!(v5.is_zero());
|
||
assert!(v6.is_zero());
|
||
|
||
// Check that the components order is correct.
|
||
m3.m11 = 11.0;
|
||
m3.m12 = 12.0;
|
||
m3.m13 = 13.0;
|
||
m3.m21 = 21.0;
|
||
m3.m22 = 22.0;
|
||
m3.m23 = 23.0;
|
||
m3.m31 = 31.0;
|
||
m3.m32 = 32.0;
|
||
m3.m33 = 33.0;
|
||
|
||
let expected_m3 = Matrix3::new(11.0, 12.0, 13.0, 21.0, 22.0, 23.0, 31.0, 32.0, 33.0);
|
||
assert_eq!(expected_m3, m3);
|
||
}
|
||
|
||
#[test]
|
||
fn kronecker() {
|
||
let a = Matrix2x3::new(11, 12, 13, 21, 22, 23);
|
||
|
||
let b = Matrix4x5::new(
|
||
110, 120, 130, 140, 150, 210, 220, 230, 240, 250, 310, 320, 330, 340, 350, 410, 420, 430,
|
||
440, 450,
|
||
);
|
||
|
||
let expected = MatrixMN::<_, U8, U15>::from_row_slice(&[
|
||
1210, 1320, 1430, 1540, 1650, 1320, 1440, 1560, 1680, 1800, 1430, 1560, 1690, 1820, 1950,
|
||
2310, 2420, 2530, 2640, 2750, 2520, 2640, 2760, 2880, 3000, 2730, 2860, 2990, 3120, 3250,
|
||
3410, 3520, 3630, 3740, 3850, 3720, 3840, 3960, 4080, 4200, 4030, 4160, 4290, 4420, 4550,
|
||
4510, 4620, 4730, 4840, 4950, 4920, 5040, 5160, 5280, 5400, 5330, 5460, 5590, 5720, 5850,
|
||
2310, 2520, 2730, 2940, 3150, 2420, 2640, 2860, 3080, 3300, 2530, 2760, 2990, 3220, 3450,
|
||
4410, 4620, 4830, 5040, 5250, 4620, 4840, 5060, 5280, 5500, 4830, 5060, 5290, 5520, 5750,
|
||
6510, 6720, 6930, 7140, 7350, 6820, 7040, 7260, 7480, 7700, 7130, 7360, 7590, 7820, 8050,
|
||
8610, 8820, 9030, 9240, 9450, 9020, 9240, 9460, 9680, 9900, 9430, 9660, 9890, 10120, 10350,
|
||
]);
|
||
|
||
let computed = a.kronecker(&b);
|
||
|
||
assert_eq!(computed, expected);
|
||
|
||
let a = Vector2::new(1, 2);
|
||
let b = Vector3::new(10, 20, 30);
|
||
let expected = Vector6::new(10, 20, 30, 20, 40, 60);
|
||
|
||
assert_eq!(a.kronecker(&b), expected);
|
||
|
||
let a = Vector2::new(1, 2);
|
||
let b = RowVector4::new(10, 20, 30, 40);
|
||
let expected = Matrix2x4::new(10, 20, 30, 40, 20, 40, 60, 80);
|
||
|
||
assert_eq!(a.kronecker(&b), expected);
|
||
}
|
||
|
||
#[test]
|
||
fn set_row_column() {
|
||
let a = Matrix4x5::new(
|
||
11, 12, 13, 14, 15, 21, 22, 23, 24, 25, 31, 32, 33, 34, 35, 41, 42, 43, 44, 45,
|
||
);
|
||
|
||
let expected1 = Matrix4x5::new(
|
||
11, 12, 13, 14, 15, 42, 43, 44, 45, 46, 31, 32, 33, 34, 35, 41, 42, 43, 44, 45,
|
||
);
|
||
|
||
let expected2 = Matrix4x5::new(
|
||
11, 12, 100, 14, 15, 42, 43, 101, 45, 46, 31, 32, 102, 34, 35, 41, 42, 103, 44, 45,
|
||
);
|
||
|
||
let row = RowVector5::new(42, 43, 44, 45, 46);
|
||
let col = Vector4::new(100, 101, 102, 103);
|
||
|
||
let mut computed = a;
|
||
|
||
computed.set_row(1, &row);
|
||
assert_eq!(expected1, computed);
|
||
|
||
computed.set_column(2, &col);
|
||
assert_eq!(expected2, computed);
|
||
}
|
||
|
||
#[test]
|
||
fn partial_clamp() {
|
||
// NOTE: from #401.
|
||
let n = Vector2::new(1.5, 0.0);
|
||
let min = Vector2::new(-75.0, -0.0);
|
||
let max = Vector2::new(75.0, 0.0);
|
||
let inter = na::partial_clamp(&n, &min, &max);
|
||
assert_eq!(*inter.unwrap(), n);
|
||
}
|
||
|
||
#[test]
|
||
fn partial_cmp() {
|
||
let a = Vector2::new(1.0, 6.0);
|
||
let b = Vector2::new(1.0, 3.0);
|
||
let c = Vector2::new(2.0, 7.0);
|
||
let d = Vector2::new(0.0, 7.0);
|
||
assert_eq!(a.partial_cmp(&a), Some(Ordering::Equal));
|
||
assert_eq!(a.partial_cmp(&b), Some(Ordering::Greater));
|
||
assert_eq!(a.partial_cmp(&c), Some(Ordering::Less));
|
||
assert_eq!(a.partial_cmp(&d), None);
|
||
}
|
||
|
||
#[test]
|
||
fn swizzle() {
|
||
let a = Vector2::new(1.0f32, 2.0);
|
||
let b = Vector3::new(1.0f32, 2.0, 3.0);
|
||
let c = Vector4::new(1.0f32, 2.0, 3.0, 4.0);
|
||
|
||
assert_eq!(a.xy(), Vector2::new(1.0, 2.0));
|
||
assert_eq!(a.yx(), Vector2::new(2.0, 1.0));
|
||
assert_eq!(a.xx(), Vector2::new(1.0, 1.0));
|
||
assert_eq!(a.yy(), Vector2::new(2.0, 2.0));
|
||
|
||
assert_eq!(a.xxx(), Vector3::new(1.0, 1.0, 1.0));
|
||
assert_eq!(a.yyy(), Vector3::new(2.0, 2.0, 2.0));
|
||
assert_eq!(a.xyx(), Vector3::new(1.0, 2.0, 1.0));
|
||
assert_eq!(a.yxy(), Vector3::new(2.0, 1.0, 2.0));
|
||
|
||
assert_eq!(b.xy(), Vector2::new(1.0, 2.0));
|
||
assert_eq!(b.yx(), Vector2::new(2.0, 1.0));
|
||
assert_eq!(b.xx(), Vector2::new(1.0, 1.0));
|
||
assert_eq!(b.yy(), Vector2::new(2.0, 2.0));
|
||
|
||
assert_eq!(b.xz(), Vector2::new(1.0, 3.0));
|
||
assert_eq!(b.zx(), Vector2::new(3.0, 1.0));
|
||
assert_eq!(b.yz(), Vector2::new(2.0, 3.0));
|
||
assert_eq!(b.zy(), Vector2::new(3.0, 2.0));
|
||
assert_eq!(b.zz(), Vector2::new(3.0, 3.0));
|
||
|
||
assert_eq!(b.xyz(), Vector3::new(1.0, 2.0, 3.0));
|
||
assert_eq!(b.xxx(), Vector3::new(1.0, 1.0, 1.0));
|
||
assert_eq!(b.yyy(), Vector3::new(2.0, 2.0, 2.0));
|
||
assert_eq!(b.zzz(), Vector3::new(3.0, 3.0, 3.0));
|
||
assert_eq!(b.zxy(), Vector3::new(3.0, 1.0, 2.0));
|
||
assert_eq!(b.zxz(), Vector3::new(3.0, 1.0, 3.0));
|
||
assert_eq!(b.zyz(), Vector3::new(3.0, 2.0, 3.0));
|
||
|
||
assert_eq!(c.xy(), Vector2::new(1.0, 2.0));
|
||
assert_eq!(c.yx(), Vector2::new(2.0, 1.0));
|
||
assert_eq!(c.xx(), Vector2::new(1.0, 1.0));
|
||
assert_eq!(c.yy(), Vector2::new(2.0, 2.0));
|
||
|
||
assert_eq!(c.xz(), Vector2::new(1.0, 3.0));
|
||
assert_eq!(c.zx(), Vector2::new(3.0, 1.0));
|
||
assert_eq!(c.yz(), Vector2::new(2.0, 3.0));
|
||
assert_eq!(c.zy(), Vector2::new(3.0, 2.0));
|
||
assert_eq!(c.zz(), Vector2::new(3.0, 3.0));
|
||
|
||
assert_eq!(c.xyz(), Vector3::new(1.0, 2.0, 3.0));
|
||
assert_eq!(c.xxx(), Vector3::new(1.0, 1.0, 1.0));
|
||
assert_eq!(c.yyy(), Vector3::new(2.0, 2.0, 2.0));
|
||
assert_eq!(c.zzz(), Vector3::new(3.0, 3.0, 3.0));
|
||
assert_eq!(c.zxy(), Vector3::new(3.0, 1.0, 2.0));
|
||
assert_eq!(c.zxz(), Vector3::new(3.0, 1.0, 3.0));
|
||
assert_eq!(c.zyz(), Vector3::new(3.0, 2.0, 3.0));
|
||
}
|
||
|
||
#[cfg(feature = "arbitrary")]
|
||
mod transposition_tests {
|
||
use super::*;
|
||
use na::Matrix4x6;
|
||
|
||
quickcheck! {
|
||
fn transpose_transpose_is_self(m: Matrix2x3<f64>) -> bool {
|
||
m.transpose().transpose() == m
|
||
}
|
||
|
||
fn transpose_mut_transpose_mut_is_self(m: Matrix3<f64>) -> bool {
|
||
let mut mm = m;
|
||
mm.transpose_mut();
|
||
mm.transpose_mut();
|
||
m == mm
|
||
}
|
||
|
||
fn transpose_transpose_is_id_dyn(m: DMatrix<f64>) -> bool {
|
||
m.transpose().transpose() == m
|
||
}
|
||
|
||
fn check_transpose_components_dyn(m: DMatrix<f64>) -> bool {
|
||
let tr = m.transpose();
|
||
let (nrows, ncols) = m.shape();
|
||
|
||
if nrows != tr.shape().1 || ncols != tr.shape().0 {
|
||
return false
|
||
}
|
||
|
||
for i in 0 .. nrows {
|
||
for j in 0 .. ncols {
|
||
if m[(i, j)] != tr[(j, i)] {
|
||
return false
|
||
}
|
||
}
|
||
}
|
||
|
||
true
|
||
}
|
||
|
||
fn tr_mul_is_transpose_then_mul(m: Matrix4x6<f64>, v: Vector4<f64>) -> bool {
|
||
relative_eq!(m.transpose() * v, m.tr_mul(&v), epsilon = 1.0e-7)
|
||
}
|
||
}
|
||
}
|
||
|
||
#[cfg(feature = "arbitrary")]
|
||
mod inversion_tests {
|
||
use super::*;
|
||
use na::Matrix1;
|
||
|
||
quickcheck! {
|
||
fn self_mul_inv_is_id_dim1(m: Matrix1<f64>) -> bool {
|
||
if let Some(im) = m.try_inverse() {
|
||
let id = Matrix1::one();
|
||
relative_eq!(im * m, id, epsilon = 1.0e-7) &&
|
||
relative_eq!(m * im, id, epsilon = 1.0e-7)
|
||
}
|
||
else {
|
||
true
|
||
}
|
||
}
|
||
|
||
fn self_mul_inv_is_id_dim2(m: Matrix2<f64>) -> bool {
|
||
if let Some(im) = m.try_inverse() {
|
||
let id = Matrix2::one();
|
||
relative_eq!(im * m, id, epsilon = 1.0e-7) &&
|
||
relative_eq!(m * im, id, epsilon = 1.0e-7)
|
||
}
|
||
else {
|
||
true
|
||
}
|
||
}
|
||
|
||
fn self_mul_inv_is_id_dim3(m: Matrix3<f64>) -> bool {
|
||
if let Some(im) = m.try_inverse() {
|
||
let id = Matrix3::one();
|
||
relative_eq!(im * m, id, epsilon = 1.0e-7) &&
|
||
relative_eq!(m * im, id, epsilon = 1.0e-7)
|
||
}
|
||
else {
|
||
true
|
||
}
|
||
}
|
||
|
||
fn self_mul_inv_is_id_dim4(m: Matrix4<f64>) -> bool {
|
||
if let Some(im) = m.try_inverse() {
|
||
let id = Matrix4::one();
|
||
relative_eq!(im * m, id, epsilon = 1.0e-7) &&
|
||
relative_eq!(m * im, id, epsilon = 1.0e-7)
|
||
}
|
||
else {
|
||
true
|
||
}
|
||
}
|
||
|
||
fn self_mul_inv_is_id_dim6(m: Matrix6<f64>) -> bool {
|
||
if let Some(im) = m.try_inverse() {
|
||
let id = Matrix6::one();
|
||
relative_eq!(im * m, id, epsilon = 1.0e-7) &&
|
||
relative_eq!(m * im, id, epsilon = 1.0e-7)
|
||
}
|
||
else {
|
||
true
|
||
}
|
||
}
|
||
}
|
||
}
|
||
|
||
#[cfg(feature = "arbitrary")]
|
||
mod normalization_tests {
|
||
use super::*;
|
||
|
||
quickcheck! {
|
||
fn normalized_vec_norm_is_one(v: Vector3<f64>) -> bool {
|
||
if let Some(nv) = v.try_normalize(1.0e-10) {
|
||
relative_eq!(nv.norm(), 1.0, epsilon = 1.0e-7)
|
||
}
|
||
else {
|
||
true
|
||
}
|
||
}
|
||
|
||
fn normalized_vec_norm_is_one_dyn(v: DVector<f64>) -> bool {
|
||
if let Some(nv) = v.try_normalize(1.0e-10) {
|
||
relative_eq!(nv.norm(), 1.0, epsilon = 1.0e-7)
|
||
}
|
||
else {
|
||
true
|
||
}
|
||
}
|
||
}
|
||
}
|
||
|
||
#[cfg(feature = "arbitrary")]
|
||
// FIXME: move this to alga ?
|
||
mod finite_dim_inner_space_tests {
|
||
use super::*;
|
||
use alga::linear::FiniteDimInnerSpace;
|
||
use std::fmt::Display;
|
||
|
||
macro_rules! finite_dim_inner_space_test(
|
||
($($Vector: ident, $orthonormal_subspace: ident, $orthonormalization: ident);* $(;)*) => {$(
|
||
quickcheck!{
|
||
fn $orthonormal_subspace(vs: Vec<$Vector<f64>>) -> bool {
|
||
let mut given_basis = vs.clone();
|
||
let given_basis_dim = $Vector::orthonormalize(&mut given_basis[..]);
|
||
let mut ortho_basis = Vec::new();
|
||
$Vector::orthonormal_subspace_basis(
|
||
&given_basis[.. given_basis_dim],
|
||
|e| { ortho_basis.push(*e); true }
|
||
);
|
||
|
||
if !is_subspace_basis(&ortho_basis[..]) {
|
||
return false;
|
||
}
|
||
|
||
for v in vs {
|
||
for b in &ortho_basis {
|
||
if !relative_eq!(v.dot(b), 0.0, epsilon = 1.0e-7) {
|
||
println!("Found dot product: {} · {} = {}", v, b, v.dot(b));
|
||
return false;
|
||
}
|
||
}
|
||
}
|
||
|
||
true
|
||
}
|
||
|
||
fn $orthonormalization(vs: Vec<$Vector<f64>>) -> bool {
|
||
let mut basis = vs.clone();
|
||
let subdim = $Vector::orthonormalize(&mut basis[..]);
|
||
|
||
if !is_subspace_basis(&basis[.. subdim]) {
|
||
return false;
|
||
}
|
||
|
||
for mut e in vs {
|
||
for b in &basis[.. subdim] {
|
||
e -= e.dot(b) * b
|
||
}
|
||
|
||
// Any element of `e` must be a linear combination of the basis elements.
|
||
if !relative_eq!(e.norm(), 0.0, epsilon = 1.0e-7) {
|
||
println!("Orthonormalization; element decomposition failure: {}", e);
|
||
println!("... the non-zero norm is: {}", e.norm());
|
||
return false;
|
||
}
|
||
}
|
||
|
||
true
|
||
}
|
||
}
|
||
)*}
|
||
);
|
||
|
||
finite_dim_inner_space_test!(
|
||
Vector1, orthonormal_subspace_basis1, orthonormalize1;
|
||
Vector2, orthonormal_subspace_basis2, orthonormalize2;
|
||
Vector3, orthonormal_subspace_basis3, orthonormalize3;
|
||
Vector4, orthonormal_subspace_basis4, orthonormalize4;
|
||
Vector5, orthonormal_subspace_basis5, orthonormalize5;
|
||
Vector6, orthonormal_subspace_basis6, orthonormalize6;
|
||
);
|
||
|
||
/*
|
||
*
|
||
* Helper functions.
|
||
*
|
||
*/
|
||
#[cfg(feature = "arbitrary")]
|
||
fn is_subspace_basis<T: FiniteDimInnerSpace<Real = f64> + Display>(vs: &[T]) -> bool {
|
||
for i in 0..vs.len() {
|
||
// Basis elements must be normalized.
|
||
if !relative_eq!(vs[i].norm(), 1.0, epsilon = 1.0e-7) {
|
||
println!("Non-zero basis element norm: {}", vs[i].norm());
|
||
return false;
|
||
}
|
||
|
||
for j in 0..i {
|
||
// Basis elements must be orthogonal.
|
||
if !relative_eq!(vs[i].dot(&vs[j]), 0.0, epsilon = 1.0e-7) {
|
||
println!(
|
||
"Non-orthogonal basis elements: {} · {} = {}",
|
||
vs[i],
|
||
vs[j],
|
||
vs[i].dot(&vs[j])
|
||
);
|
||
return false;
|
||
}
|
||
}
|
||
}
|
||
|
||
true
|
||
}
|
||
}
|