The column may not be normalized if the `factor` is on a scale of 1e-40.
Possibly, f32 just runs out of precision.
There is likely a better solution to the problem.
@ -200,16 +248,21 @@ This project adheres to [Semantic Versioning](https://semver.org/).
cargo features.
cargo features.
## [0.28.0]
## [0.28.0]
### Added
### Added
- Implement `Hash` for `Transform`.
- Implement `Hash` for `Transform`.
- Implement `Borrow` and `BorrowMut` for contiguous slices.
- Implement `Borrow` and `BorrowMut` for contiguous slices.
### Modified
### Modified
- The `OPoint<T, D>` type has been added. It takes the dimension number as a type-level integer (e.g. `Const<3>`) instead
- The `OPoint<T, D>` type has been added. It takes the dimension number as a type-level integer (e.g. `Const<3>`)
instead
of a const-generic. The type `Point<T, const D: usize>` is now an alias for `OPoint`. This changes doesn't affect any
of a const-generic. The type `Point<T, const D: usize>` is now an alias for `OPoint`. This changes doesn't affect any
of the existing code using `Point`. However, it will allow the use `OPoint` in a generic context where the dimension
of the existing code using `Point`. However, it will allow the use `OPoint` in a generic context where the dimension
cannot be easily expressed as a const-generic (because of the current limitation of const-generics in Rust).
cannot be easily expressed as a const-generic (because of the current limitation of const-generics in Rust).
- Several clippy warnings were fixed. This results in some method signature changes (e.g. taking `self` instead of `&self`)
- Several clippy warnings were fixed. This results in some method signature changes (e.g. taking `self` instead
of `&self`)
but this should not have any practical infulances on existing codebase.
but this should not have any practical infulances on existing codebase.
- The `Point::new` constructors are no longer const-fn. This is due to some limitations in const-fn
- The `Point::new` constructors are no longer const-fn. This is due to some limitations in const-fn
not allowing custom trait-bounds. Use the `point!` macro instead to build points in const environments.
not allowing custom trait-bounds. Use the `point!` macro instead to build points in const environments.
@ -217,30 +270,38 @@ This project adheres to [Semantic Versioning](https://semver.org/).
- Methods returning `Result<(), ()>` now return `bool` instead.
- Methods returning `Result<(), ()>` now return `bool` instead.
### Fixed
### Fixed
- Fixed a potential unsoundess issue when converting a mutable slice to a `&mut[T]`.
- Fixed a potential unsoundess issue when converting a mutable slice to a `&mut[T]`.
## [0.27.1]
## [0.27.1]
### Fixed
### Fixed
- Fixed a bug in the conversion from `glam::Vec2` or `glam::DVec2` to `Isometry2`.
- Fixed a bug in the conversion from `glam::Vec2` or `glam::DVec2` to `Isometry2`.
## [0.27.0]
## [0.27.0]
This removes the `convert-glam` and `convert-glam-unchecked` optional features.
This removes the `convert-glam` and `convert-glam-unchecked` optional features.
Instead, this adds the `convert-glam013`, `convert-glam014`, and `convert-glam015` optional features for
Instead, this adds the `convert-glam013`, `convert-glam014`, and `convert-glam015` optional features for
conversions targeting the versions 0.13, 0.14, and 0.15 of `glam`.
conversions targeting the versions 0.13, 0.14, and 0.15 of `glam`.
### Added
### Added
- Add macros `matrix!`, `dmatrix!`, `vector!`, `dvector!`, `point!` for constructing matrices/vectors/points in a
- Add macros `matrix!`, `dmatrix!`, `vector!`, `dvector!`, `point!` for constructing matrices/vectors/points in a
more convenient way. See [#886](https://github.com/dimforge/nalgebra/pull/886) and [#899](https://github.com/dimforge/nalgebra/pull/899).
more convenient way. See [#886](https://github.com/dimforge/nalgebra/pull/886)
and [#899](https://github.com/dimforge/nalgebra/pull/899).
- Add `CooMatrix::reserve` to `nalgebra-sparse`.
- Add `CooMatrix::reserve` to `nalgebra-sparse`.
- Add basic support for serialization using `rkyv`. Can be enabled with the features `rkyv-serialize` or
- Add basic support for serialization using `rkyv`. Can be enabled with the features `rkyv-serialize` or
`rkyv-serialize-no-std`.
`rkyv-serialize-no-std`.
### Fixed
### Fixed
- Fixed a potential unsoundness issue after deserializing an invalid `DVector` using `serde`.
- Fixed a potential unsoundness issue after deserializing an invalid `DVector` using `serde`.
## [0.26.2]
## [0.26.2]
### Added
### Added
- Conversion from an array `[T; D]` to an isometry `Isometry<T, _, D>` (as a translation).
- Conversion from an array `[T; D]` to an isometry `Isometry<T, _, D>` (as a translation).
- Conversion from a static vector `SVector<T; D>` to an isometry `Isometry<T, _, D>` (as a translation).
- Conversion from a static vector `SVector<T; D>` to an isometry `Isometry<T, _, D>` (as a translation).
- Conversion from a point `Point<T; D>` to an isometry `Isometry<T, _, D>` (as a translation).
- Conversion from a point `Point<T; D>` to an isometry `Isometry<T, _, D>` (as a translation).
@ -250,14 +311,17 @@ conversions targeting the versions 0.13, 0.14, and 0.15 of `glam`.
- Conversion of a glam type `Vec2/3/4` from/to a `Translation2/3/4`.
- Conversion of a glam type `Vec2/3/4` from/to a `Translation2/3/4`.
## [0.26.1]
## [0.26.1]
Fix a regression introduced in 0.26.0 preventing `DVector` from being serialized with `serde`.
Fix a regression introduced in 0.26.0 preventing `DVector` from being serialized with `serde`.
## [0.26.0]
## [0.26.0]
This release integrates `min-const-generics` to nalgebra. See
This release integrates `min-const-generics` to nalgebra. See
[our blog post](https://www.dimforge.com/blog/2021/04/12/integrating-const-generics-to-nalgebra)
[our blog post](https://www.dimforge.com/blog/2021/04/12/integrating-const-generics-to-nalgebra)
for details about this release.
for details about this release.
### Added
### Added
- Add type aliases for unit vector, e.g., `UnitVector3`.
- Add type aliases for unit vector, e.g., `UnitVector3`.
- Add a `pow` and `pow_mut` function to square matrices.
- Add a `pow` and `pow_mut` function to square matrices.
- Add `Cholesky::determinant` to compute the determinant of a matrix decomposed
- Add `Cholesky::determinant` to compute the determinant of a matrix decomposed
@ -265,8 +329,8 @@ for details about this release.
- Add the `serde-serialize-no-std` feature to enable serialization of static matrices/vectors
- Add the `serde-serialize-no-std` feature to enable serialization of static matrices/vectors
with serde, but without requiring `std`.
with serde, but without requiring `std`.
### Modified
### Modified
- The `serde` crate isn't enabled by default now. Enable the `serde-serialize` or the
- The `serde` crate isn't enabled by default now. Enable the `serde-serialize` or the
`serde-serialize-no-std` features instead.
`serde-serialize-no-std` features instead.
- The `Const<const D: usize>` type has been introduced to represent dimensions known
- The `Const<const D: usize>` type has been introduced to represent dimensions known
@ -288,16 +352,22 @@ for details about this release.
constructor is also a `const fn` now.
constructor is also a `const fn` now.
## [0.25.4]
## [0.25.4]
### Fixed
### Fixed
- Fix a compilation error when only the `serde-serialize` feature is enabled.
- Fix a compilation error when only the `serde-serialize` feature is enabled.
## [0.25.3]
## [0.25.3]
### Added
### Added
- The `Vector::simd_cap_magnitude` method to cap the magnitude of the vector with
- The `Vector::simd_cap_magnitude` method to cap the magnitude of the vector with
SIMD components.
SIMD components.
## [0.25.2]
## [0.25.2]
### Added
### Added
- A `convert-glam` cargo feature to enable implementations of `From` traits to convert
- A `convert-glam` cargo feature to enable implementations of `From` traits to convert
between `glam` types and `nalgebra` types.
between `glam` types and `nalgebra` types.
- A `convert-glam-unchecked` cargo feature to enable some extra `glam`/`nalgebra` conversions that may
- A `convert-glam-unchecked` cargo feature to enable some extra `glam`/`nalgebra` conversions that may
@ -308,56 +378,70 @@ for details about this release.
type of the components of a given entity. Example: `vector.cast::<f32>()`.
type of the components of a given entity. Example: `vector.cast::<f32>()`.
## [0.25.1]
## [0.25.1]
This release replaces the version 0.25.0 which has been yanked. The 0.25.0 version
This release replaces the version 0.25.0 which has been yanked. The 0.25.0 version
added significant complication to build `nalgebra` targeting a `#[no-std]` platform
added significant complication to build `nalgebra` targeting a `#[no-std]` platform
not supported by `rand`.
not supported by `rand`.
The `rand` dependency is now optional (and disabled by default). You may enable it with:
The `rand` dependency is now optional (and disabled by default). You may enable it with:
- The `rand-no-std` cargo feature when targeting a `#[no-std]` environment.
- The `rand-no-std` cargo feature when targeting a `#[no-std]` environment.
- The `rand` cargo feature when targeting a `std` environment.
- The `rand` cargo feature when targeting a `std` environment.
## [0.25.0] - Yanked
## [0.25.0] - Yanked
This updates all the dependencies of nalgebra to their latest version, including:
This updates all the dependencies of nalgebra to their latest version, including:
- rand 0.8
- rand 0.8
- proptest 1.0
- proptest 1.0
- simba 0.4
- simba 0.4
### New crate: nalgebra-sparse
### New crate: nalgebra-sparse
Alongside this release of `nalgebra`, we are releasing `nalgebra-sparse`: a crate dedicated to sparse matrix
Alongside this release of `nalgebra`, we are releasing `nalgebra-sparse`: a crate dedicated to sparse matrix
computation with `nalgebra`. The `sparse` module of `nalgebra`itself still exists for backward compatibility,
computation with `nalgebra`. The `sparse` module of `nalgebra`itself still exists for backward compatibility,
but it will be deprecated soon in favor of the `nalgebra-sparse` crate.
but it will be deprecated soon in favor of the `nalgebra-sparse` crate.
### Added
### Added
* Add `UnitDualQuaternion`, a dual-quaternion with unit magnitude which can be used as an isometry transformation.
* Add `UnitDualQuaternion`, a dual-quaternion with unit magnitude which can be used as an isometry transformation.
* Add `UDU::new()` and `matrix.udu()` to compute the UDU factorization of a matrix.
* Add `UDU::new()` and `matrix.udu()` to compute the UDU factorization of a matrix.
* Add `ColPivQR::new()` and `matrix.col_piv_qr()` to compute the QR decomposition with column pivoting of a matrix.
* Add `ColPivQR::new()` and `matrix.col_piv_qr()` to compute the QR decomposition with column pivoting of a matrix.
* Add `from_basis_unchecked` to all the rotation types. This builds a rotation from a set of basis vectors (representing the columns of the corresponding rotation matrix).
* Add `from_basis_unchecked` to all the rotation types. This builds a rotation from a set of basis vectors (representing
the columns of the corresponding rotation matrix).
* Add `Matrix::cap_magnitude` to cap the magnitude of a vector.
* Add `Matrix::cap_magnitude` to cap the magnitude of a vector.
* Add `UnitQuaternion::append_axisangle_linearized` to approximately append a rotation represented as an axis-angle to a rotation represented as an unit quaternion.
* Add `UnitQuaternion::append_axisangle_linearized` to approximately append a rotation represented as an axis-angle to a
rotation represented as an unit quaternion.
* Mark the iterators on matrix components as `DoubleEndedIter`.
* Mark the iterators on matrix components as `DoubleEndedIter`.
* Re-export `simba::simd::SimdValue` at the root of the `nalgebra` crate.
* Re-export `simba::simd::SimdValue` at the root of the `nalgebra` crate.
## [0.24.0]
## [0.24.0]
### Added
### Added
* The `DualQuaternion` type. It is still work-in-progress, but the basics are here:
* The `DualQuaternion` type. It is still work-in-progress, but the basics are here:
creation from its real and dual part, multiplication of two dual quaternions,
creation from its real and dual part, multiplication of two dual quaternions,
and normalization.
and normalization.
### Removed
### Removed
* There is no blanket `impl<T> PartialEq for Unit<T>` anymore. Instead, it is
* There is no blanket `impl<T> PartialEq for Unit<T>` anymore. Instead, it is
implemented specifically for `UnitComplex`, `UnitQuaternion` and `Unit<Vector>`.
implemented specifically for `UnitComplex`, `UnitQuaternion` and `Unit<Vector>`.
## [0.23.2]
## [0.23.2]
In this release, we improved the documentation of some of the geometric types
In this release, we improved the documentation of some of the geometric types
by applying changes similar to what we did in the version 0.23.1 for matrices.
by applying changes similar to what we did in the version 0.23.1 for matrices.
### Added
### Added
* The `Isometry::inv_mul` method which is a more efficient way of doing
* The `Isometry::inv_mul` method which is a more efficient way of doing
`isometry1.inverse() * isometry2`.
`isometry1.inverse() * isometry2`.
## [0.23.1]
## [0.23.1]
In this release we improved the documentation of the matrix and vector types by:
In this release we improved the documentation of the matrix and vector types by:
- Grouping `impl` bocks logically, adding a title comment to these impl blocks.
- Grouping `impl` bocks logically, adding a title comment to these impl blocks.
- Reference these impl blocks docs at the top of the documentation page for `Matrix`.
- Reference these impl blocks docs at the top of the documentation page for `Matrix`.
- Reduce the depth of type aliasing. Now all vector and matrix types are aliases of `Matrix`
- Reduce the depth of type aliasing. Now all vector and matrix types are aliases of `Matrix`
@ -366,13 +450,17 @@ In this release we improved the documentation of the matrix and vector types by:
## [0.23.0]
## [0.23.0]
### Added
### Added
* The `.inverse_transform_unit_vector(v)` was added to `Rotation2/3`, `Isometry2/3`, `UnitQuaternion`, and `UnitComplex`.
* The `.inverse_transform_unit_vector(v)` was added to `Rotation2/3`, `Isometry2/3`, `UnitQuaternion`,
and `UnitComplex`.
It applies the corresponding rotation to a unit vector `Unit<Vector2/3>`.
It applies the corresponding rotation to a unit vector `Unit<Vector2/3>`.
* The `Point.map(f)` and `Point.apply(f)` to apply a function to each component of the point, similarly to `Vector.map(f)`
* The `Point.map(f)` and `Point.apply(f)` to apply a function to each component of the point, similarly
to `Vector.map(f)`
and `Vector.apply(f)`.
and `Vector.apply(f)`.
* The `Quaternion::from([N; 4])` conversion to build a quaternion from an array of four elements.
* The `Quaternion::from([N; 4])` conversion to build a quaternion from an array of four elements.
* The `Isometry::from(Translation)` conversion to build an isometry from a translation.
* The `Isometry::from(Translation)` conversion to build an isometry from a translation.
* The `Vector::ith_axis(i)` which build a unit vector, e.g., `Unit<Vector3<f32>>` with its i-th component set to 1.0, and the
* The `Vector::ith_axis(i)` which build a unit vector, e.g., `Unit<Vector3<f32>>` with its i-th component set to 1.0,
and the
others set to zero.
others set to zero.
* The `Isometry.lerp_slerp` and `Isometry.try_lerp_slerp` methods to interpolate between two isometries using linear
* The `Isometry.lerp_slerp` and `Isometry.try_lerp_slerp` methods to interpolate between two isometries using linear
interpolation for the translational part, and spherical interpolation for the rotational part.
interpolation for the translational part, and spherical interpolation for the rotational part.
@ -380,12 +468,14 @@ In this release we improved the documentation of the matrix and vector types by:
spherical interpolation.
spherical interpolation.
## [0.22.0]
## [0.22.0]
In this release, we are using the new version 0.2 of simba. One major change of that version is that the
In this release, we are using the new version 0.2 of simba. One major change of that version is that the
use of `libm` is now opt-in when building targeting `no-std` environment. If you are using floating-point
use of `libm` is now opt-in when building targeting `no-std` environment. If you are using floating-point
operations with nalgebra in a `no-std` environment, you will need to enable the new `libm` feature
operations with nalgebra in a `no-std` environment, you will need to enable the new `libm` feature
of nalgebra for your code to compile again.
of nalgebra for your code to compile again.
### Added
### Added
* The `libm` feature that enables `libm` when building for `no-std` environment.
* The `libm` feature that enables `libm` when building for `no-std` environment.
* The `libm-force` feature that enables `libm` even when building for a not `no-std` environment.
* The `libm-force` feature that enables `libm` even when building for a not `no-std` environment.
* `Cholesky::new_unchecked` which build a Cholesky decomposition without checking that its input is
* `Cholesky::new_unchecked` which build a Cholesky decomposition without checking that its input is
@ -396,6 +486,7 @@ of nalgebra for your code to compile again.
* The `Vector::ith(i, x)` that builds a vector filled with zeros except for the `i`-th component set to `x`.
* The `Vector::ith(i, x)` that builds a vector filled with zeros except for the `i`-th component set to `x`.
## [0.21.0]
## [0.21.0]
In this release, we are no longer relying on traits from the __alga__ crate for our generic code.
In this release, we are no longer relying on traits from the __alga__ crate for our generic code.
Instead, we use traits from the new [simba](https://crates.io/crates/simba) crate which are both
Instead, we use traits from the new [simba](https://crates.io/crates/simba) crate which are both
simpler, and allow for significant optimizations like AoSoA SIMD.
simpler, and allow for significant optimizations like AoSoA SIMD.
@ -404,75 +495,98 @@ Refer to the [monthly dimforge blogpost](https://www.dimforge.org/blog/2020/04/0
for details about this switch and its benefits.
for details about this switch and its benefits.
### Added
### Added
* It is now possible to use SIMD types like `simba::f32x4` as scalar types for nalgebra's matrices and
* It is now possible to use SIMD types like `simba::f32x4` as scalar types for nalgebra's matrices and
geometric types.
geometric types.
### Modified
### Modified
* Use of traits like `alga::general::{RealField, ComplexField}` have now been replaced by
* Use of traits like `alga::general::{RealField, ComplexField}` have now been replaced by
`simba::scalar::{RealField, ComplexField}`.
`simba::scalar::{RealField, ComplexField}`.
* The implementation of traits from the __alga__ crate (and well as the dependency to _alga__) are now
* The implementation of traits from the __alga__ crate (and well as the dependency to _alga__) are now
omitted unless the `alga` cargo feature is activated.
omitted unless the `alga` cargo feature is activated.
### Removed
### Removed
* The `Neg` unary operator is no longer implemented for `UnitComplex` and `UnitQuaternion`. This caused
* The `Neg` unary operator is no longer implemented for `UnitComplex` and `UnitQuaternion`. This caused
hard-to-track errors when we mistakenly write, e.g., `-q * v` instead of `-(q * v)`.
hard-to-track errors when we mistakenly write, e.g., `-q * v` instead of `-(q * v)`.
* The `na::convert_unchecked` is no longer marked as unsafe.
* The `na::convert_unchecked` is no longer marked as unsafe.
## [0.20.0]
## [0.20.0]
### Added
### Added
* `cholesky.rank_one_update(...)` which performs a rank-one update on the cholesky decomposition of a matrix.
* `cholesky.rank_one_update(...)` which performs a rank-one update on the cholesky decomposition of a matrix.
* `From<&Matrix>` is now implemented for matrix slices.
* `From<&Matrix>` is now implemented for matrix slices.
* `.try_set_magnitude(...)` which sets the magnitude of a vector, while keeping its direction.
* `.try_set_magnitude(...)` which sets the magnitude of a vector, while keeping its direction.
* Implementations of `From` and `Into` for the conversion between matrix slices and standard (`&[N]` `&mut [N]`) slices.
* Implementations of `From` and `Into` for the conversion between matrix slices and standard (`&[N]` `&mut [N]`) slices.
### Modified
### Modified
* We started some major changes in order to allow non-Copy types to be used as scalar types inside of matrices/vectors.
* We started some major changes in order to allow non-Copy types to be used as scalar types inside of matrices/vectors.
## [0.19.0]
## [0.19.0]
### Added
### Added
* `.remove_rows_at` and `remove_columns_at` which removes a set of rows or columns (specified by indices) from a matrix.
* `.remove_rows_at` and `remove_columns_at` which removes a set of rows or columns (specified by indices) from a matrix.
* Several formatting traits have been implemented for all matrices/vectors: `LowerExp`, `UpperExp`, `Octal`, `LowerHex`,
* Several formatting traits have been implemented for all matrices/vectors: `LowerExp`, `UpperExp`, `Octal`, `LowerHex`,
`UpperHex`, `Binary`, `Pointer`.
`UpperHex`, `Binary`, `Pointer`.
* `UnitQuaternion::quaternions_mean(...)` which computes the mean rotation of a set of unit quaternions. This implements
* `UnitQuaternion::quaternions_mean(...)` which computes the mean rotation of a set of unit quaternions. This implements
the algorithm from _Oshman, Yaakov, and Avishy Carmi, "Attitude estimation from vector observations using a genetic-algorithm-embedded quaternion particle filter."
the algorithm from _Oshman, Yaakov, and Avishy Carmi, "Attitude estimation from vector observations using a
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
