pub struct ColliderPosition(pub Isometry<f32>);
Expand description
The position of a collider.
Tuple Fields§
§0: Isometry<f32>
Implementations§
Methods from Deref<Target = Isometry<f32>>§
Sourcepub fn inverse(&self) -> Isometry<T, R, D>
pub fn inverse(&self) -> Isometry<T, R, D>
Inverts self
.
§Example
let iso = Isometry2::new(Vector2::new(1.0, 2.0), f32::consts::FRAC_PI_2);
let inv = iso.inverse();
let pt = Point2::new(1.0, 2.0);
assert_eq!(inv * (iso * pt), pt);
Sourcepub fn inverse_mut(&mut self)
pub fn inverse_mut(&mut self)
Inverts self
in-place.
§Example
let mut iso = Isometry2::new(Vector2::new(1.0, 2.0), f32::consts::FRAC_PI_2);
let pt = Point2::new(1.0, 2.0);
let transformed_pt = iso * pt;
iso.inverse_mut();
assert_eq!(iso * transformed_pt, pt);
Sourcepub fn inv_mul(&self, rhs: &Isometry<T, R, D>) -> Isometry<T, R, D>
pub fn inv_mul(&self, rhs: &Isometry<T, R, D>) -> Isometry<T, R, D>
Computes self.inverse() * rhs
in a more efficient way.
§Example
let mut iso1 = Isometry2::new(Vector2::new(1.0, 2.0), f32::consts::FRAC_PI_2);
let mut iso2 = Isometry2::new(Vector2::new(10.0, 20.0), f32::consts::FRAC_PI_4);
assert_eq!(iso1.inverse() * iso2, iso1.inv_mul(&iso2));
Sourcepub fn append_translation_mut(&mut self, t: &Translation<T, D>)
pub fn append_translation_mut(&mut self, t: &Translation<T, D>)
Appends to self
the given translation in-place.
§Example
let mut iso = Isometry2::new(Vector2::new(1.0, 2.0), f32::consts::FRAC_PI_2);
let tra = Translation2::new(3.0, 4.0);
// Same as `iso = tra * iso`.
iso.append_translation_mut(&tra);
assert_eq!(iso.translation, Translation2::new(4.0, 6.0));
Sourcepub fn append_rotation_mut(&mut self, r: &R)
pub fn append_rotation_mut(&mut self, r: &R)
Appends to self
the given rotation in-place.
§Example
let mut iso = Isometry2::new(Vector2::new(1.0, 2.0), f32::consts::PI / 6.0);
let rot = UnitComplex::new(f32::consts::PI / 2.0);
// Same as `iso = rot * iso`.
iso.append_rotation_mut(&rot);
assert_relative_eq!(iso, Isometry2::new(Vector2::new(-2.0, 1.0), f32::consts::PI * 2.0 / 3.0), epsilon = 1.0e-6);
Sourcepub fn append_rotation_wrt_point_mut(&mut self, r: &R, p: &OPoint<T, Const<D>>)
pub fn append_rotation_wrt_point_mut(&mut self, r: &R, p: &OPoint<T, Const<D>>)
Appends in-place to self
a rotation centered at the point p
, i.e., the rotation that
lets p
invariant.
§Example
let mut iso = Isometry2::new(Vector2::new(1.0, 2.0), f32::consts::FRAC_PI_2);
let rot = UnitComplex::new(f32::consts::FRAC_PI_2);
let pt = Point2::new(1.0, 0.0);
iso.append_rotation_wrt_point_mut(&rot, &pt);
assert_relative_eq!(iso * pt, Point2::new(-2.0, 0.0), epsilon = 1.0e-6);
Sourcepub fn append_rotation_wrt_center_mut(&mut self, r: &R)
pub fn append_rotation_wrt_center_mut(&mut self, r: &R)
Appends in-place to self
a rotation centered at the point with coordinates
self.translation
.
§Example
let mut iso = Isometry2::new(Vector2::new(1.0, 2.0), f32::consts::FRAC_PI_2);
let rot = UnitComplex::new(f32::consts::FRAC_PI_2);
iso.append_rotation_wrt_center_mut(&rot);
// The translation part should not have changed.
assert_eq!(iso.translation.vector, Vector2::new(1.0, 2.0));
assert_eq!(iso.rotation, UnitComplex::new(f32::consts::PI));
Sourcepub fn transform_point(&self, pt: &OPoint<T, Const<D>>) -> OPoint<T, Const<D>>
pub fn transform_point(&self, pt: &OPoint<T, Const<D>>) -> OPoint<T, Const<D>>
Transform the given point by this isometry.
This is the same as the multiplication self * pt
.
§Example
let tra = Translation3::new(0.0, 0.0, 3.0);
let rot = UnitQuaternion::from_scaled_axis(Vector3::y() * f32::consts::FRAC_PI_2);
let iso = Isometry3::from_parts(tra, rot);
let transformed_point = iso.transform_point(&Point3::new(1.0, 2.0, 3.0));
assert_relative_eq!(transformed_point, Point3::new(3.0, 2.0, 2.0), epsilon = 1.0e-6);
Sourcepub fn transform_vector(
&self,
v: &Matrix<T, Const<D>, Const<1>, ArrayStorage<T, D, 1>>,
) -> Matrix<T, Const<D>, Const<1>, ArrayStorage<T, D, 1>>
pub fn transform_vector( &self, v: &Matrix<T, Const<D>, Const<1>, ArrayStorage<T, D, 1>>, ) -> Matrix<T, Const<D>, Const<1>, ArrayStorage<T, D, 1>>
Transform the given vector by this isometry, ignoring the translation component of the isometry.
This is the same as the multiplication self * v
.
§Example
let tra = Translation3::new(0.0, 0.0, 3.0);
let rot = UnitQuaternion::from_scaled_axis(Vector3::y() * f32::consts::FRAC_PI_2);
let iso = Isometry3::from_parts(tra, rot);
let transformed_point = iso.transform_vector(&Vector3::new(1.0, 2.0, 3.0));
assert_relative_eq!(transformed_point, Vector3::new(3.0, 2.0, -1.0), epsilon = 1.0e-6);
Sourcepub fn inverse_transform_point(
&self,
pt: &OPoint<T, Const<D>>,
) -> OPoint<T, Const<D>>
pub fn inverse_transform_point( &self, pt: &OPoint<T, Const<D>>, ) -> OPoint<T, Const<D>>
Transform the given point by the inverse of this isometry. This may be less expensive than computing the entire isometry inverse and then transforming the point.
§Example
let tra = Translation3::new(0.0, 0.0, 3.0);
let rot = UnitQuaternion::from_scaled_axis(Vector3::y() * f32::consts::FRAC_PI_2);
let iso = Isometry3::from_parts(tra, rot);
let transformed_point = iso.inverse_transform_point(&Point3::new(1.0, 2.0, 3.0));
assert_relative_eq!(transformed_point, Point3::new(0.0, 2.0, 1.0), epsilon = 1.0e-6);
Sourcepub fn inverse_transform_vector(
&self,
v: &Matrix<T, Const<D>, Const<1>, ArrayStorage<T, D, 1>>,
) -> Matrix<T, Const<D>, Const<1>, ArrayStorage<T, D, 1>>
pub fn inverse_transform_vector( &self, v: &Matrix<T, Const<D>, Const<1>, ArrayStorage<T, D, 1>>, ) -> Matrix<T, Const<D>, Const<1>, ArrayStorage<T, D, 1>>
Transform the given vector by the inverse of this isometry, ignoring the translation component of the isometry. This may be less expensive than computing the entire isometry inverse and then transforming the point.
§Example
let tra = Translation3::new(0.0, 0.0, 3.0);
let rot = UnitQuaternion::from_scaled_axis(Vector3::y() * f32::consts::FRAC_PI_2);
let iso = Isometry3::from_parts(tra, rot);
let transformed_point = iso.inverse_transform_vector(&Vector3::new(1.0, 2.0, 3.0));
assert_relative_eq!(transformed_point, Vector3::new(-3.0, 2.0, 1.0), epsilon = 1.0e-6);
Sourcepub fn inverse_transform_unit_vector(
&self,
v: &Unit<Matrix<T, Const<D>, Const<1>, ArrayStorage<T, D, 1>>>,
) -> Unit<Matrix<T, Const<D>, Const<1>, ArrayStorage<T, D, 1>>>
pub fn inverse_transform_unit_vector( &self, v: &Unit<Matrix<T, Const<D>, Const<1>, ArrayStorage<T, D, 1>>>, ) -> Unit<Matrix<T, Const<D>, Const<1>, ArrayStorage<T, D, 1>>>
Transform the given unit vector by the inverse of this isometry, ignoring the translation component of the isometry. This may be less expensive than computing the entire isometry inverse and then transforming the point.
§Example
let tra = Translation3::new(0.0, 0.0, 3.0);
let rot = UnitQuaternion::from_scaled_axis(Vector3::z() * f32::consts::FRAC_PI_2);
let iso = Isometry3::from_parts(tra, rot);
let transformed_point = iso.inverse_transform_unit_vector(&Vector3::x_axis());
assert_relative_eq!(transformed_point, -Vector3::y_axis(), epsilon = 1.0e-6);
Sourcepub fn to_homogeneous(
&self,
) -> Matrix<T, <Const<D> as DimNameAdd<Const<1>>>::Output, <Const<D> as DimNameAdd<Const<1>>>::Output, <DefaultAllocator as Allocator<<Const<D> as DimNameAdd<Const<1>>>::Output, <Const<D> as DimNameAdd<Const<1>>>::Output>>::Buffer<T>>where
Const<D>: DimNameAdd<Const<1>>,
R: SubsetOf<Matrix<T, <Const<D> as DimNameAdd<Const<1>>>::Output, <Const<D> as DimNameAdd<Const<1>>>::Output, <DefaultAllocator as Allocator<<Const<D> as DimNameAdd<Const<1>>>::Output, <Const<D> as DimNameAdd<Const<1>>>::Output>>::Buffer<T>>>,
DefaultAllocator: Allocator<<Const<D> as DimNameAdd<Const<1>>>::Output, <Const<D> as DimNameAdd<Const<1>>>::Output>,
pub fn to_homogeneous(
&self,
) -> Matrix<T, <Const<D> as DimNameAdd<Const<1>>>::Output, <Const<D> as DimNameAdd<Const<1>>>::Output, <DefaultAllocator as Allocator<<Const<D> as DimNameAdd<Const<1>>>::Output, <Const<D> as DimNameAdd<Const<1>>>::Output>>::Buffer<T>>where
Const<D>: DimNameAdd<Const<1>>,
R: SubsetOf<Matrix<T, <Const<D> as DimNameAdd<Const<1>>>::Output, <Const<D> as DimNameAdd<Const<1>>>::Output, <DefaultAllocator as Allocator<<Const<D> as DimNameAdd<Const<1>>>::Output, <Const<D> as DimNameAdd<Const<1>>>::Output>>::Buffer<T>>>,
DefaultAllocator: Allocator<<Const<D> as DimNameAdd<Const<1>>>::Output, <Const<D> as DimNameAdd<Const<1>>>::Output>,
Converts this isometry into its equivalent homogeneous transformation matrix.
This is the same as self.to_matrix()
.
§Example
let iso = Isometry2::new(Vector2::new(10.0, 20.0), f32::consts::FRAC_PI_6);
let expected = Matrix3::new(0.8660254, -0.5, 10.0,
0.5, 0.8660254, 20.0,
0.0, 0.0, 1.0);
assert_relative_eq!(iso.to_homogeneous(), expected, epsilon = 1.0e-6);
Sourcepub fn to_matrix(
&self,
) -> Matrix<T, <Const<D> as DimNameAdd<Const<1>>>::Output, <Const<D> as DimNameAdd<Const<1>>>::Output, <DefaultAllocator as Allocator<<Const<D> as DimNameAdd<Const<1>>>::Output, <Const<D> as DimNameAdd<Const<1>>>::Output>>::Buffer<T>>where
Const<D>: DimNameAdd<Const<1>>,
R: SubsetOf<Matrix<T, <Const<D> as DimNameAdd<Const<1>>>::Output, <Const<D> as DimNameAdd<Const<1>>>::Output, <DefaultAllocator as Allocator<<Const<D> as DimNameAdd<Const<1>>>::Output, <Const<D> as DimNameAdd<Const<1>>>::Output>>::Buffer<T>>>,
DefaultAllocator: Allocator<<Const<D> as DimNameAdd<Const<1>>>::Output, <Const<D> as DimNameAdd<Const<1>>>::Output>,
pub fn to_matrix(
&self,
) -> Matrix<T, <Const<D> as DimNameAdd<Const<1>>>::Output, <Const<D> as DimNameAdd<Const<1>>>::Output, <DefaultAllocator as Allocator<<Const<D> as DimNameAdd<Const<1>>>::Output, <Const<D> as DimNameAdd<Const<1>>>::Output>>::Buffer<T>>where
Const<D>: DimNameAdd<Const<1>>,
R: SubsetOf<Matrix<T, <Const<D> as DimNameAdd<Const<1>>>::Output, <Const<D> as DimNameAdd<Const<1>>>::Output, <DefaultAllocator as Allocator<<Const<D> as DimNameAdd<Const<1>>>::Output, <Const<D> as DimNameAdd<Const<1>>>::Output>>::Buffer<T>>>,
DefaultAllocator: Allocator<<Const<D> as DimNameAdd<Const<1>>>::Output, <Const<D> as DimNameAdd<Const<1>>>::Output>,
Converts this isometry into its equivalent homogeneous transformation matrix.
This is the same as self.to_homogeneous()
.
§Example
let iso = Isometry2::new(Vector2::new(10.0, 20.0), f32::consts::FRAC_PI_6);
let expected = Matrix3::new(0.8660254, -0.5, 10.0,
0.5, 0.8660254, 20.0,
0.0, 0.0, 1.0);
assert_relative_eq!(iso.to_matrix(), expected, epsilon = 1.0e-6);
Sourcepub fn lerp_slerp(
&self,
other: &Isometry<T, Unit<Quaternion<T>>, 3>,
t: T,
) -> Isometry<T, Unit<Quaternion<T>>, 3>where
T: RealField,
pub fn lerp_slerp(
&self,
other: &Isometry<T, Unit<Quaternion<T>>, 3>,
t: T,
) -> Isometry<T, Unit<Quaternion<T>>, 3>where
T: RealField,
Interpolates between two isometries using a linear interpolation for the translation part, and a spherical interpolation for the rotation part.
Panics if the angle between both rotations is 180 degrees (in which case the interpolation
is not well-defined). Use .try_lerp_slerp
instead to avoid the panic.
§Examples:
let t1 = Translation3::new(1.0, 2.0, 3.0);
let t2 = Translation3::new(4.0, 8.0, 12.0);
let q1 = UnitQuaternion::from_euler_angles(std::f32::consts::FRAC_PI_4, 0.0, 0.0);
let q2 = UnitQuaternion::from_euler_angles(-std::f32::consts::PI, 0.0, 0.0);
let iso1 = Isometry3::from_parts(t1, q1);
let iso2 = Isometry3::from_parts(t2, q2);
let iso3 = iso1.lerp_slerp(&iso2, 1.0 / 3.0);
assert_eq!(iso3.translation.vector, Vector3::new(2.0, 4.0, 6.0));
assert_eq!(iso3.rotation.euler_angles(), (std::f32::consts::FRAC_PI_2, 0.0, 0.0));
Sourcepub fn try_lerp_slerp(
&self,
other: &Isometry<T, Unit<Quaternion<T>>, 3>,
t: T,
epsilon: T,
) -> Option<Isometry<T, Unit<Quaternion<T>>, 3>>where
T: RealField,
pub fn try_lerp_slerp(
&self,
other: &Isometry<T, Unit<Quaternion<T>>, 3>,
t: T,
epsilon: T,
) -> Option<Isometry<T, Unit<Quaternion<T>>, 3>>where
T: RealField,
Attempts to interpolate between two isometries using a linear interpolation for the translation part, and a spherical interpolation for the rotation part.
Returns None
if the angle between both rotations is 180 degrees (in which case the interpolation
is not well-defined).
§Examples:
let t1 = Translation3::new(1.0, 2.0, 3.0);
let t2 = Translation3::new(4.0, 8.0, 12.0);
let q1 = UnitQuaternion::from_euler_angles(std::f32::consts::FRAC_PI_4, 0.0, 0.0);
let q2 = UnitQuaternion::from_euler_angles(-std::f32::consts::PI, 0.0, 0.0);
let iso1 = Isometry3::from_parts(t1, q1);
let iso2 = Isometry3::from_parts(t2, q2);
let iso3 = iso1.lerp_slerp(&iso2, 1.0 / 3.0);
assert_eq!(iso3.translation.vector, Vector3::new(2.0, 4.0, 6.0));
assert_eq!(iso3.rotation.euler_angles(), (std::f32::consts::FRAC_PI_2, 0.0, 0.0));
Sourcepub fn lerp_slerp(
&self,
other: &Isometry<T, Rotation<T, 3>, 3>,
t: T,
) -> Isometry<T, Rotation<T, 3>, 3>where
T: RealField,
pub fn lerp_slerp(
&self,
other: &Isometry<T, Rotation<T, 3>, 3>,
t: T,
) -> Isometry<T, Rotation<T, 3>, 3>where
T: RealField,
Interpolates between two isometries using a linear interpolation for the translation part, and a spherical interpolation for the rotation part.
Panics if the angle between both rotations is 180 degrees (in which case the interpolation
is not well-defined). Use .try_lerp_slerp
instead to avoid the panic.
§Examples:
let t1 = Translation3::new(1.0, 2.0, 3.0);
let t2 = Translation3::new(4.0, 8.0, 12.0);
let q1 = Rotation3::from_euler_angles(std::f32::consts::FRAC_PI_4, 0.0, 0.0);
let q2 = Rotation3::from_euler_angles(-std::f32::consts::PI, 0.0, 0.0);
let iso1 = IsometryMatrix3::from_parts(t1, q1);
let iso2 = IsometryMatrix3::from_parts(t2, q2);
let iso3 = iso1.lerp_slerp(&iso2, 1.0 / 3.0);
assert_eq!(iso3.translation.vector, Vector3::new(2.0, 4.0, 6.0));
assert_eq!(iso3.rotation.euler_angles(), (std::f32::consts::FRAC_PI_2, 0.0, 0.0));
Sourcepub fn try_lerp_slerp(
&self,
other: &Isometry<T, Rotation<T, 3>, 3>,
t: T,
epsilon: T,
) -> Option<Isometry<T, Rotation<T, 3>, 3>>where
T: RealField,
pub fn try_lerp_slerp(
&self,
other: &Isometry<T, Rotation<T, 3>, 3>,
t: T,
epsilon: T,
) -> Option<Isometry<T, Rotation<T, 3>, 3>>where
T: RealField,
Attempts to interpolate between two isometries using a linear interpolation for the translation part, and a spherical interpolation for the rotation part.
Returns None
if the angle between both rotations is 180 degrees (in which case the interpolation
is not well-defined).
§Examples:
let t1 = Translation3::new(1.0, 2.0, 3.0);
let t2 = Translation3::new(4.0, 8.0, 12.0);
let q1 = Rotation3::from_euler_angles(std::f32::consts::FRAC_PI_4, 0.0, 0.0);
let q2 = Rotation3::from_euler_angles(-std::f32::consts::PI, 0.0, 0.0);
let iso1 = IsometryMatrix3::from_parts(t1, q1);
let iso2 = IsometryMatrix3::from_parts(t2, q2);
let iso3 = iso1.lerp_slerp(&iso2, 1.0 / 3.0);
assert_eq!(iso3.translation.vector, Vector3::new(2.0, 4.0, 6.0));
assert_eq!(iso3.rotation.euler_angles(), (std::f32::consts::FRAC_PI_2, 0.0, 0.0));
Sourcepub fn lerp_slerp(
&self,
other: &Isometry<T, Unit<Complex<T>>, 2>,
t: T,
) -> Isometry<T, Unit<Complex<T>>, 2>where
T: RealField,
pub fn lerp_slerp(
&self,
other: &Isometry<T, Unit<Complex<T>>, 2>,
t: T,
) -> Isometry<T, Unit<Complex<T>>, 2>where
T: RealField,
Interpolates between two isometries using a linear interpolation for the translation part, and a spherical interpolation for the rotation part.
Panics if the angle between both rotations is 180 degrees (in which case the interpolation
is not well-defined). Use .try_lerp_slerp
instead to avoid the panic.
§Examples:
let t1 = Translation2::new(1.0, 2.0);
let t2 = Translation2::new(4.0, 8.0);
let q1 = UnitComplex::new(std::f32::consts::FRAC_PI_4);
let q2 = UnitComplex::new(-std::f32::consts::PI);
let iso1 = Isometry2::from_parts(t1, q1);
let iso2 = Isometry2::from_parts(t2, q2);
let iso3 = iso1.lerp_slerp(&iso2, 1.0 / 3.0);
assert_eq!(iso3.translation.vector, Vector2::new(2.0, 4.0));
assert_relative_eq!(iso3.rotation.angle(), std::f32::consts::FRAC_PI_2);
Sourcepub fn lerp_slerp(
&self,
other: &Isometry<T, Rotation<T, 2>, 2>,
t: T,
) -> Isometry<T, Rotation<T, 2>, 2>where
T: RealField,
pub fn lerp_slerp(
&self,
other: &Isometry<T, Rotation<T, 2>, 2>,
t: T,
) -> Isometry<T, Rotation<T, 2>, 2>where
T: RealField,
Interpolates between two isometries using a linear interpolation for the translation part, and a spherical interpolation for the rotation part.
Panics if the angle between both rotations is 180 degrees (in which case the interpolation
is not well-defined). Use .try_lerp_slerp
instead to avoid the panic.
§Examples:
let t1 = Translation2::new(1.0, 2.0);
let t2 = Translation2::new(4.0, 8.0);
let q1 = Rotation2::new(std::f32::consts::FRAC_PI_4);
let q2 = Rotation2::new(-std::f32::consts::PI);
let iso1 = IsometryMatrix2::from_parts(t1, q1);
let iso2 = IsometryMatrix2::from_parts(t2, q2);
let iso3 = iso1.lerp_slerp(&iso2, 1.0 / 3.0);
assert_eq!(iso3.translation.vector, Vector2::new(2.0, 4.0));
assert_relative_eq!(iso3.rotation.angle(), std::f32::consts::FRAC_PI_2);
Trait Implementations§
Source§impl AsMut<Isometry<f32, Unit<Quaternion<f32>>, 3>> for ColliderPosition
impl AsMut<Isometry<f32, Unit<Quaternion<f32>>, 3>> for ColliderPosition
Source§impl AsRef<Isometry<f32, Unit<Quaternion<f32>>, 3>> for ColliderPosition
impl AsRef<Isometry<f32, Unit<Quaternion<f32>>, 3>> for ColliderPosition
Source§impl Clone for ColliderPosition
impl Clone for ColliderPosition
Source§fn clone(&self) -> ColliderPosition
fn clone(&self) -> ColliderPosition
1.0.0 · Source§fn clone_from(&mut self, source: &Self)
fn clone_from(&mut self, source: &Self)
source
. Read moreSource§impl Debug for ColliderPosition
impl Debug for ColliderPosition
Source§impl Default for ColliderPosition
impl Default for ColliderPosition
Source§impl Deref for ColliderPosition
impl Deref for ColliderPosition
Source§impl DerefMut for ColliderPosition
impl DerefMut for ColliderPosition
Source§impl<T> From<T> for ColliderPosition
impl<T> From<T> for ColliderPosition
Source§impl PartialEq for ColliderPosition
impl PartialEq for ColliderPosition
impl Copy for ColliderPosition
impl StructuralPartialEq for ColliderPosition
Auto Trait Implementations§
impl Freeze for ColliderPosition
impl RefUnwindSafe for ColliderPosition
impl Send for ColliderPosition
impl Sync for ColliderPosition
impl Unpin for ColliderPosition
impl UnwindSafe for ColliderPosition
Blanket Implementations§
Source§impl<T> BorrowMut<T> for Twhere
T: ?Sized,
impl<T> BorrowMut<T> for Twhere
T: ?Sized,
Source§fn borrow_mut(&mut self) -> &mut T
fn borrow_mut(&mut self) -> &mut T
Source§impl<T> CloneToUninit for Twhere
T: Clone,
impl<T> CloneToUninit for Twhere
T: Clone,
Source§impl<T> Downcast for Twhere
T: Any,
impl<T> Downcast for Twhere
T: Any,
Source§fn into_any(self: Box<T>) -> Box<dyn Any>
fn into_any(self: Box<T>) -> Box<dyn Any>
Box<dyn Trait>
(where Trait: Downcast
) to Box<dyn Any>
. Box<dyn Any>
can
then be further downcast
into Box<ConcreteType>
where ConcreteType
implements Trait
.Source§fn into_any_rc(self: Rc<T>) -> Rc<dyn Any>
fn into_any_rc(self: Rc<T>) -> Rc<dyn Any>
Rc<Trait>
(where Trait: Downcast
) to Rc<Any>
. Rc<Any>
can then be
further downcast
into Rc<ConcreteType>
where ConcreteType
implements Trait
.Source§fn as_any(&self) -> &(dyn Any + 'static)
fn as_any(&self) -> &(dyn Any + 'static)
&Trait
(where Trait: Downcast
) to &Any
. This is needed since Rust cannot
generate &Any
’s vtable from &Trait
’s.Source§fn as_any_mut(&mut self) -> &mut (dyn Any + 'static)
fn as_any_mut(&mut self) -> &mut (dyn Any + 'static)
&mut Trait
(where Trait: Downcast
) to &Any
. This is needed since Rust cannot
generate &mut Any
’s vtable from &mut Trait
’s.Source§impl<T> DowncastSync for T
impl<T> DowncastSync for T
Source§impl<T> IntoEither for T
impl<T> IntoEither for T
Source§fn into_either(self, into_left: bool) -> Either<Self, Self>
fn into_either(self, into_left: bool) -> Either<Self, Self>
self
into a Left
variant of Either<Self, Self>
if into_left
is true
.
Converts self
into a Right
variant of Either<Self, Self>
otherwise. Read moreSource§fn into_either_with<F>(self, into_left: F) -> Either<Self, Self>
fn into_either_with<F>(self, into_left: F) -> Either<Self, Self>
self
into a Left
variant of Either<Self, Self>
if into_left(&self)
returns true
.
Converts self
into a Right
variant of Either<Self, Self>
otherwise. Read moreSource§impl<T> Pointable for T
impl<T> Pointable for T
Source§impl<SS, SP> SupersetOf<SS> for SPwhere
SS: SubsetOf<SP>,
impl<SS, SP> SupersetOf<SS> for SPwhere
SS: SubsetOf<SP>,
Source§fn to_subset(&self) -> Option<SS>
fn to_subset(&self) -> Option<SS>
self
from the equivalent element of its
superset. Read moreSource§fn is_in_subset(&self) -> bool
fn is_in_subset(&self) -> bool
self
is actually part of its subset T
(and can be converted to it).Source§fn to_subset_unchecked(&self) -> SS
fn to_subset_unchecked(&self) -> SS
self.to_subset
but without any property checks. Always succeeds.Source§fn from_subset(element: &SS) -> SP
fn from_subset(element: &SS) -> SP
self
to the equivalent element of its superset.