Implement the single-trait approach.

CHANGELOG.md

@@ -4,6 +4,13 @@ documented here.
 
 This project adheres to [Semantic Versioning](https://semver.org/).
 
+## [0.29.0]
+### Modified
+- The closure given to `apply`, `zip_apply`, `zip_zip_apply` must now modify the
+  first argument inplace, instead of returning a new value. This makes these
+  methods more versatile, and avoid useless clones when using non-Copy scalar
+  types.
+
 ## [0.28.0]
 ### Added
 - Implement `Hash` for `Transform`.

Cargo.toml

@@ -119,4 +119,4 @@ lto = true
 
 [package.metadata.docs.rs]
 # Enable certain features when building docs for docs.rs
-features = [ "proptest-support", "compare", "macros" ]
+features = [ "proptest-support", "compare", "macros", "rand" ]

benches/core/matrix.rs

@@ -1,7 +1,4 @@
-use na::{
-    Const, DMatrix, DVector, Dynamic, Matrix2, Matrix3, Matrix4, OMatrix, Vector2, Vector3,
-    Vector4, U10,
-};
+use na::{DMatrix, DVector, Matrix2, Matrix3, Matrix4, OMatrix, Vector2, Vector3, Vector4, U10};
 use rand::Rng;
 use rand_isaac::IsaacRng;
 use std::ops::{Add, Div, Mul, Sub};
@@ -189,15 +186,15 @@ fn axpy(bench: &mut criterion::Criterion) {
 fn tr_mul_to(bench: &mut criterion::Criterion) {
     let a = DMatrix::<f64>::new_random(1000, 1000);
     let b = DVector::<f64>::new_random(1000);
-    let mut c = DVector::new_uninitialized_generic(Dynamic::new(1000), Const::<1>);
+    let mut c = DVector::from_element(1000, 0.0);
 
     bench.bench_function("tr_mul_to", move |bh| bh.iter(|| a.tr_mul_to(&b, &mut c)));
 }
 
 fn mat_mul_mat(bench: &mut criterion::Criterion) {
     let a = DMatrix::<f64>::new_random(100, 100);
     let b = DMatrix::<f64>::new_random(100, 100);
-    let mut ab = DMatrix::new_uninitialized_generic(Dynamic::new(100), Dynamic::new(100));
+    let mut ab = DMatrix::<f64>::from_element(100, 100, 0.0);
 
     bench.bench_function("mat_mul_mat", move |bh| {
         bh.iter(|| {

clippy.toml

@@ -0,0 +1,2 @@
+too-many-arguments-threshold = 8
+type-complexity-threshold = 675

nalgebra-glm/src/aliases.rs

@@ -320,7 +320,7 @@ pub type DMat4x4 = Matrix4<f64>;
 pub type Mat2 = Matrix2<f32>;
 /// A 2x2 matrix with `f32` components.
 pub type Mat2x2 = Matrix2<f32>;
-/// A 2x2 matrix with `f32` components.
+/// A 2x3 matrix with `f32` components.
 pub type Mat2x3 = Matrix2x3<f32>;
 /// A 2x4 matrix with `f32` components.
 pub type Mat2x4 = Matrix2x4<f32>;

nalgebra-glm/src/lib.rs

@@ -110,6 +110,16 @@
    and keep in mind it is possible to convert, e.g., an `Isometry3` to a `Mat4` and vice-versa (see the [conversions section](#conversions)).
 */
 
+#![deny(
+    nonstandard_style,
+    unused,
+    missing_docs,
+    rust_2018_idioms,
+    rust_2018_compatibility,
+    future_incompatible,
+    missing_copy_implementations,
+    missing_debug_implementations
+)]
 #![doc(html_favicon_url = "https://nalgebra.org/img/favicon.ico")]
 #![cfg_attr(not(feature = "std"), no_std)]
 

nalgebra-lapack/src/cholesky.rs

@@ -6,7 +6,7 @@ use num_complex::Complex;
 
 use na::allocator::Allocator;
 use na::dimension::Dim;
-use na::storage::Storage;
+use na::storage::RawStorage;
 use na::{DefaultAllocator, Matrix, OMatrix, Scalar};
 
 use lapack;
@@ -24,17 +24,17 @@ use lapack;
          OMatrix<T, D, D>: Deserialize<'de>"))
 )]
 #[derive(Clone, Debug)]
-pub struct Cholesky<T, D: Dim>
+pub struct Cholesky<T: Scalar, D: Dim>
 where
     DefaultAllocator: Allocator<T, D, D>,
 {
     l: OMatrix<T, D, D>,
 }
 
-impl<T: Copy, D: Dim> Copy for Cholesky<T, D>
+impl<T: Scalar + Copy, D: Dim> Copy for Cholesky<T, D>
 where
     DefaultAllocator: Allocator<T, D, D>,
-    Owned<T, D, D>: Copy,
+    OMatrix<T, D, D>: Copy,
 {
 }
 
@@ -104,7 +104,7 @@ where
         b: &Matrix<T, R2, C2, S2>,
     ) -> Option<OMatrix<T, R2, C2>>
     where
-        S2: Storage<T, R2, C2>,
+        S2: RawStorage<T, R2, C2>,
         DefaultAllocator: Allocator<T, R2, C2>,
     {
         let mut res = b.clone_owned();

nalgebra-lapack/src/eigen.rs

@@ -1,5 +1,3 @@
-use std::fmt;
-
 #[cfg(feature = "serde-serialize")]
 use serde::{Deserialize, Serialize};
 
@@ -11,7 +9,7 @@ use simba::scalar::RealField;
 use crate::ComplexHelper;
 use na::allocator::Allocator;
 use na::dimension::{Const, Dim};
-use na::storage::Storage;
+use na::storage::RawStorage;
 use na::{DefaultAllocator, Matrix, OMatrix, OVector, Scalar};
 
 use lapack;
@@ -34,7 +32,8 @@ use lapack;
          OMatrix<T, D, D>: Deserialize<'de>")
     )
 )]
-pub struct Eigen<T, D: Dim>
+#[derive(Clone, Debug)]
+pub struct Eigen<T: Scalar, D: Dim>
 where
     DefaultAllocator: Allocator<T, D> + Allocator<T, D, D>,
 {
@@ -46,44 +45,14 @@ where
     pub left_eigenvectors: Option<OMatrix<T, D, D>>,
 }
 
-impl<T: Copy, D: Dim> Copy for Eigen<T, D>
+impl<T: Scalar + Copy, D: Dim> Copy for Eigen<T, D>
 where
     DefaultAllocator: Allocator<T, D> + Allocator<T, D, D>,
     OVector<T, D>: Copy,
     OMatrix<T, D, D>: Copy,
 {
 }
 
-impl<T: Clone, D: Dim> Clone for Eigen<T, D>
-where
-    DefaultAllocator: Allocator<T, D> + Allocator<T, D, D>,
-    OVector<T, D>: Clone,
-    OMatrix<T, D, D>: Clone,
-{
-    fn clone(&self) -> Self {
-        Self {
-            eigenvalues: self.eigenvalues.clone(),
-            eigenvectors: self.eigenvectors.clone(),
-            left_eigenvectors: self.left_eigenvectors.clone(),
-        }
-    }
-}
-
-impl<T: fmt::Debug, D: Dim> fmt::Debug for Eigen<T, D>
-where
-    DefaultAllocator: Allocator<T, D> + Allocator<T, D, D>,
-    OVector<T, D>: fmt::Debug,
-    OMatrix<T, D, D>: fmt::Debug,
-{
-    fn fmt(&self, f: &mut fmt::Formatter) -> fmt::Result {
-        f.debug_struct("Eigen")
-            .field("eigenvalues", &self.eigenvalues)
-            .field("eigenvectors", &self.eigenvectors)
-            .field("left_eigenvectors", &self.left_eigenvectors)
-            .finish()
-    }
-}
-
 impl<T: EigenScalar + RealField, D: Dim> Eigen<T, D>
 where
     DefaultAllocator: Allocator<T, D, D> + Allocator<T, D>,
@@ -104,13 +73,11 @@ where
         let ljob = if left_eigenvectors { b'V' } else { b'T' };
         let rjob = if eigenvectors { b'V' } else { b'T' };
 
-        let (nrows, ncols) = m.data.shape();
+        let (nrows, ncols) = m.shape_generic();
         let n = nrows.value();
 
         let lda = n as i32;
 
-        // IMPORTANT TODO: this is still UB.
-
         let mut wr = unsafe { Matrix::new_uninitialized_generic(nrows, Const::<1>).assume_init() };
         // TODO: Tap into the workspace.
         let mut wi = unsafe { Matrix::new_uninitialized_generic(nrows, Const::<1>).assume_init() };
@@ -275,7 +242,7 @@ where
             "Unable to compute the eigenvalue decomposition of a non-square matrix."
         );
 
-        let nrows = m.data.shape().0;
+        let nrows = m.shape_generic().0;
         let n = nrows.value();
 
         let lda = n as i32;

nalgebra-lapack/src/hessenberg.rs

@@ -4,7 +4,7 @@ use num_complex::Complex;
 use crate::ComplexHelper;
 use na::allocator::Allocator;
 use na::dimension::{Const, DimDiff, DimSub, U1};
-use na::storage::Storage;
+use na::storage::RawStorage;
 use na::{DefaultAllocator, Matrix, OMatrix, OVector, Scalar};
 
 use lapack;
@@ -48,7 +48,7 @@ where
 {
     /// Computes the hessenberg decomposition of the matrix `m`.
     pub fn new(mut m: OMatrix<T, D, D>) -> Self {
-        let nrows = m.data.shape().0;
+        let nrows = m.shape_generic().0;
         let n = nrows.value() as i32;
 
         assert!(
@@ -60,7 +60,6 @@ where
             "Unable to compute the hessenberg decomposition of an empty matrix."
         );
 
-        // IMPORTANT TODO: this is still UB.
         let mut tau = unsafe {
             Matrix::new_uninitialized_generic(nrows.sub(Const::<1>), Const::<1>).assume_init()
         };
@@ -85,7 +84,7 @@ where
         );
         lapack_panic!(info);
 
-        Self { h: m, tau: tau }
+        Self { h: m, tau }
     }
 
     /// Computes the hessenberg matrix of this decomposition.

nalgebra-lapack/src/lib.rs

@@ -140,7 +140,6 @@ impl ComplexHelper for Complex<f64> {
     }
 }
 
-// This is UB.
 unsafe fn uninitialized_vec<T: Copy>(n: usize) -> Vec<T> {
     let mut res = Vec::new();
     res.reserve_exact(n);

nalgebra-lapack/src/lu.rs

@@ -4,7 +4,7 @@ use num_complex::Complex;
 use crate::ComplexHelper;
 use na::allocator::Allocator;
 use na::dimension::{Const, Dim, DimMin, DimMinimum};
-use na::storage::Storage;
+use na::storage::RawStorage;
 use na::{DefaultAllocator, Matrix, OMatrix, OVector, Scalar};
 
 use lapack;
@@ -61,7 +61,7 @@ where
 {
     /// Computes the LU decomposition with partial (row) pivoting of `matrix`.
     pub fn new(mut m: OMatrix<T, R, C>) -> Self {
-        let (nrows, ncols) = m.data.shape();
+        let (nrows, ncols) = m.shape_generic();
         let min_nrows_ncols = nrows.min(ncols);
         let nrows = nrows.value() as i32;
         let ncols = ncols.value() as i32;
@@ -87,7 +87,7 @@ where
     #[inline]
     #[must_use]
     pub fn l(&self) -> OMatrix<T, R, DimMinimum<R, C>> {
-        let (nrows, ncols) = self.lu.data.shape();
+        let (nrows, ncols) = self.lu.shape_generic();
         let mut res = self.lu.columns_generic(0, nrows.min(ncols)).into_owned();
 
         res.fill_upper_triangle(Zero::zero(), 1);
@@ -100,7 +100,7 @@ where
     #[inline]
     #[must_use]
     pub fn u(&self) -> OMatrix<T, DimMinimum<R, C>, C> {
-        let (nrows, ncols) = self.lu.data.shape();
+        let (nrows, ncols) = self.lu.shape_generic();
         let mut res = self.lu.rows_generic(0, nrows.min(ncols)).into_owned();
 
         res.fill_lower_triangle(Zero::zero(), 1);
@@ -115,7 +115,7 @@ where
     #[inline]
     #[must_use]
     pub fn p(&self) -> OMatrix<T, R, R> {
-        let (dim, _) = self.lu.data.shape();
+        let (dim, _) = self.lu.shape_generic();
         let mut id = Matrix::identity_generic(dim, dim);
         self.permute(&mut id);
 
@@ -191,7 +191,7 @@ where
         b: &Matrix<T, R2, C2, S2>,
     ) -> Option<OMatrix<T, R2, C2>>
     where
-        S2: Storage<T, R2, C2>,
+        S2: RawStorage<T, R2, C2>,
         DefaultAllocator: Allocator<T, R2, C2> + Allocator<i32, R2>,
     {
         let mut res = b.clone_owned();
@@ -209,7 +209,7 @@ where
         b: &Matrix<T, R2, C2, S2>,
     ) -> Option<OMatrix<T, R2, C2>>
     where
-        S2: Storage<T, R2, C2>,
+        S2: RawStorage<T, R2, C2>,
         DefaultAllocator: Allocator<T, R2, C2> + Allocator<i32, R2>,
     {
         let mut res = b.clone_owned();
@@ -227,7 +227,7 @@ where
         b: &Matrix<T, R2, C2, S2>,
     ) -> Option<OMatrix<T, R2, C2>>
     where
-        S2: Storage<T, R2, C2>,
+        S2: RawStorage<T, R2, C2>,
         DefaultAllocator: Allocator<T, R2, C2> + Allocator<i32, R2>,
     {
         let mut res = b.clone_owned();

nalgebra-lapack/src/qr.rs

@@ -7,7 +7,7 @@ use num_complex::Complex;
 use crate::ComplexHelper;
 use na::allocator::Allocator;
 use na::dimension::{Const, Dim, DimMin, DimMinimum};
-use na::storage::Storage;
+use na::storage::RawStorage;
 use na::{DefaultAllocator, Matrix, OMatrix, OVector, Scalar};
 
 use lapack;
@@ -54,14 +54,15 @@ where
 {
     /// Computes the QR decomposition of the matrix `m`.
     pub fn new(mut m: OMatrix<T, R, C>) -> Self {
-        let (nrows, ncols) = m.data.shape();
+        let (nrows, ncols) = m.shape_generic();
 
         let mut info = 0;
-        let mut tau =
-            unsafe { Matrix::new_uninitialized_generic(nrows.min(ncols), U1).assume_init() };
+        let mut tau = unsafe {
+            Matrix::new_uninitialized_generic(nrows.min(ncols), Const::<1>).assume_init()
+        };
 
         if nrows.value() == 0 || ncols.value() == 0 {
-            return Self { qr: m, tau: tau };
+            return Self { qr: m, tau };
         }
 
         let lwork = T::xgeqrf_work_size(
@@ -86,14 +87,14 @@ where
             &mut info,
         );
 
-        Self { qr: m, tau: tau }
+        Self { qr: m, tau }
     }
 
     /// Retrieves the upper trapezoidal submatrix `R` of this decomposition.
     #[inline]
     #[must_use]
     pub fn r(&self) -> OMatrix<T, DimMinimum<R, C>, C> {
-        let (nrows, ncols) = self.qr.data.shape();
+        let (nrows, ncols) = self.qr.shape_generic();
         self.qr.rows_generic(0, nrows.min(ncols)).upper_triangle()
     }
 }
@@ -119,7 +120,7 @@ where
     #[inline]
     #[must_use]
     pub fn q(&self) -> OMatrix<T, R, DimMinimum<R, C>> {
-        let (nrows, ncols) = self.qr.data.shape();
+        let (nrows, ncols) = self.qr.shape_generic();
         let min_nrows_ncols = nrows.min(ncols);
 
         if min_nrows_ncols.value() == 0 {