gsFunction< T > Class Template Reference

Detailed Description

template<class T>
class gismo::gsFunction< T >

A function \(f:\mathbb{R}^n\rightarrow\mathbb{R}^m\) from a n-dimensional domain to an m-dimensional image.

Implementations of gsFunction must at the very least implement the evaluation function gsFunction::eval_into(). It is also recommended to specify the source and target dimensions by overriding gsFunction::domainDim() and gsFunction::targetDim().

The functions for the derivatives may either be overridden or left as the default implementations, which use finite differences.

Evaluation members

All evaluation functions take a matrix u as an argument which specifies where the function should be evaluated. This matrix should have d rows, and every column specifies one point of the domain at which the function should be evaluated.

Here is an overview over the different evaluation procedures available:

Name of procedure	Evaluate what
eval(u)	value
deriv(u)	first derivative(s)
deriv2(u)	second derivative(s)

All evaluation functions also provide a version suffixed with _into which takes a matrix reference as an additional output parameter into which the result will be stored.

Template Parameters

T	arithmetic type

Inheritance diagram for gsFunction< T >:

Collaboration diagram for gsFunction< T >:

Public Types
typedef memory::shared_ptr < gsFunction >	Ptr
	Shared pointer for gsFunction.
typedef memory::unique_ptr < gsFunction >	uPtr
	Unique pointer for gsFunction.

Public Member Functions
gsMatrix< index_t >	active (const gsMatrix< T > &u) const
	Returns the indices of active (nonzero) functions at points u, as a list of indices. More...
void	active_into (const gsMatrix< T > &u, gsMatrix< index_t > &result) const
	Indices of active (non-zero) function(s) for each point. More...
const gsBasis< T > &	basis (const index_t k) const
	Helper which casts and returns the k-th piece of this function set as a gsBasis.
uPtr	clone ()
	Clone methode. Produceds a deep copy inside a uPtr.
virtual void	compute (const gsMatrix< T > &in, gsFuncData< T > &out) const
	Computes function data. More...
virtual void	computeMap (gsMapData< T > &InOut) const
	Computes map function data. More...
gsFuncCoordinate< T >	coord (const index_t c) const
	Returns the scalar function giving the i-th coordinate of this function.
gsMatrix< T >	deriv (const gsMatrix< T > &u) const
	Evaluate the derivatives,. More...
gsMatrix< T >	deriv2 (const gsMatrix< T > &u) const
	Evaluates the second derivatives of active (i.e., non-zero) functions at points u. More...
virtual T	distanceL2 (gsFunction< T > const &) const
	Computes the L2-distance between this function and the field and a function func.
virtual short_t	domainDim () const =0
	Dimension of the (source) domain. More...
gsMatrix< T >	eval (const gsMatrix< T > &u) const
	Evaluate the function,. More...
std::vector< gsMatrix< T > >	evalAllDers (const gsMatrix< T > &u, int n, bool sameElement=false) const
	Evaluate all derivatives upto order n,. More...
virtual void	evalAllDers_into (const gsMatrix< T > &u, int n, std::vector< gsMatrix< T > > &result, bool sameElement=false) const
	Evaluate the nonzero functions and their derivatives up to order n at points u into result. More...
const gsFunction< T > &	function (const index_t k) const
	Helper which casts and returns the k-th piece of this function set as a gsFunction.
virtual void	invertPoints (const gsMatrix< T > &points, gsMatrix< T > &result, const T accuracy=1e-6, const bool useInitialPoint=false) const
int	newtonRaphson (const gsVector< T > &value, gsVector< T > &arg, bool withSupport=true, const T accuracy=1e-6, int max_loop=100, T damping_factor=1) const
virtual index_t	nPieces () const
	Number of pieces in the domain of definition.
virtual gsMatrix< T >	parameterCenter () const
	Returns a "central" point inside inside the parameter domain.
gsMatrix< T >	parameterCenter (const boxCorner &bc) const
	Get coordinates of the boxCorner bc in the parameter domain.
gsMatrix< T >	parameterCenter (const boxSide &bs) const
	Get coordinates of the midpoint of the boxSide bs in the parameter domain.
virtual const gsFunction &	piece (const index_t k) const
	Returns the piece(s) of the function(s) at subdomain k.
virtual std::ostream &	print (std::ostream &os) const
	Prints the object as a string.
void	recoverPoints (gsMatrix< T > &xyz, gsMatrix< T > &uv, index_t k, const T accuracy=1e-6) const
index_t	size () const
	size More...
virtual short_t	targetDim () const
	Dimension of the target space. More...
Evaluation functions
These functions allow one to evaluate the function as well as its derivatives at one or more points in the parameter space. See also Evaluation members.
virtual void	eval_into (const gsMatrix< T > &u, gsMatrix< T > &result) const =0
	Evaluate the function at points u into result. More...
virtual void	eval_component_into (const gsMatrix< T > &u, const index_t comp, gsMatrix< T > &result) const
	Evaluate the function for component comp in the target dimension at points u into result.
virtual void	deriv_into (const gsMatrix< T > &u, gsMatrix< T > &result) const
	Evaluate derivatives of the function \(f:\mathbb{R}^n\rightarrow\mathbb{R}^m\) at points u into result. More...
virtual void	jacobian_into (const gsMatrix< T > &u, gsMatrix< T > &result) const
	Computes for each point u a block of result containing the Jacobian matrix.
void	div_into (const gsMatrix< T > &u, gsMatrix< T > &result) const
	Computes for each point u a block of result containing the divergence matrix.
gsMatrix< T >	jacobian (const gsMatrix< T > &u) const
virtual void	deriv2_into (const gsMatrix< T > &u, gsMatrix< T > &result) const
	Evaluate second derivatives of the function at points u into result. More...
virtual void	hessian_into (const gsMatrix< T > &u, gsMatrix< T > &result, index_t coord=0) const
virtual gsMatrix< T >	hessian (const gsMatrix< T > &u, index_t coord=0) const
virtual gsMatrix< T >	laplacian (const gsMatrix< T > &u) const
	Evaluate the Laplacian at points u. More...

Member Function Documentation

gsMatrix<index_t> active ( const gsMatrix< T > &  u ) const

inlineinherited

Returns the indices of active (nonzero) functions at points u, as a list of indices.

See Also

active_into()

void active_into ( const gsMatrix< T > &  u, gsMatrix< index_t > &  result  ) const

inlinevirtual

Indices of active (non-zero) function(s) for each point.

The columns are sorted in increasing order, if on a point there are less active then the number of rows in the result matrix (some other point has more actives) then the rest of the column is filled with 0s.

Parameters

u
result

Reimplemented from gsFunctionSet< T >.

void compute ( const gsMatrix< T > &  in, gsFuncData< T > &  out  ) const

virtualinherited

Computes function data.

This function evaluates the functions and their derivatives at the points in and writes them in the corresponding fields of out. Which field to write (and what to compute) is controlled by the out.flags (see also gsFuncData).

The input points in are expected to be compatible with the implementation/representation of the function, i.e. they should be points inside the domain of definitition of the function

Parameters

[in]	in
[out]	out

Reimplemented in gsGeometry< T >, and gsConstantFunction< T >.

void computeMap ( gsMapData< T > &  InOut ) const

virtual

Computes map function data.

This function evaluates the functions and their derivatives at the points InOut.points and writes them in the corresponding fields of InOut. Which field to write (and what to compute) is controlled by the InOut.flags (see also gsMapData). This is intended for parametrizations only and it works on functions sets of cardinality 1 only.

Parameters

[in,out]

InOut

gsMatrix< T > deriv ( const gsMatrix< T > &  u ) const

inherited

Evaluate the derivatives,.

See Also

deriv_into()

gsMatrix< T > deriv2 ( const gsMatrix< T > &  u ) const

inherited

Evaluates the second derivatives of active (i.e., non-zero) functions at points u.

See documentation for deriv2_into() (the one without input parameter coefs) for details.

See Also

deriv2_into()

Parameters

[in]

u

Evaluation points in columns.

Returns

For every column of u, a column containing the second derivatives. See documentation for deriv2_into() (the one without input parameter coefs) for details.

void deriv2_into ( const gsMatrix< T > &  u, gsMatrix< T > &  result  ) const

virtual

Evaluate second derivatives of the function at points u into result.

Let n be the dimension of the source space ( n = domainDim() ).
Let m be the dimension of the image/target space ( m = targetDim() ).
Let N denote the number of evaluation points.

Parameters

[in]	u	gsMatrix of size n x N, where each column of u represents one evaluation point.
[out]	result	gsMatrix of size (S*m) x N, where S=n*(n+1)/2. Each column in result corresponds to one point (i.e., one column in u) and contains the following values (for n=3, m=3): \( (\partial_{xx} f^{(1)}, \partial_{yy} f^{(1)}, \partial_{zz} f^{(1)}, \partial_{xy} f^{(1)}, \partial_{xz} f^{(1)}, \partial_{yz} f^{(1)}, \partial_{xx} f^{(2)},\ldots,\partial_{yz} f^{(3)} )^T\)

Warning

By default uses central finite differences with h=0.00001! One must override this function in derived classes to get proper results.

Reimplemented from gsFunctionSet< T >.

Reimplemented in gsGeometry< T >, gsFunctionExpr< T >, gsConstantFunction< T >, gsPreCICEFunction< T >, gsAffineFunction< T >, gsMappedSingleSpline< d, T >, gsGeometryTransform< T >, and gsFuncCoordinate< T >.

void deriv_into ( const gsMatrix< T > &  u, gsMatrix< T > &  result  ) const

virtual

Evaluate derivatives of the function \(f:\mathbb{R}^n\rightarrow\mathbb{R}^m\) at points u into result.

Let n be the dimension of the source space ( n = domainDim() ).
Let m be the dimension of the image/target space ( m = targetDim() ).
Let N denote the number of evaluation points.

Let \( f:\mathbb R^2 \rightarrow \mathbb R^3 \), i.e., \( f(x,y) = ( f^{(1)}(x,y), f^{(2)}(x,y), f^{(3)}(x,y) )^T\),
and let \( u = ( u_1, \ldots, u_N) = ( (x_1,y_1)^T, \ldots, (x_N, y_N)^T )\).
Then, result is of the form

\[ \left[ \begin{array}{cccc} \partial_x f^{(1)}(u_1) & \partial_x f^{(1)}(u_2) & \ldots & \partial_x f^{(1)}(u_N) \\ \partial_y f^{(1)}(u_1) & \partial_y f^{(1)}(u_2) & \ldots & \partial_y f^{(1)}(u_N) \\ \partial_x f^{(2)}(u_1) & \partial_x f^{(2)}(u_2) & \ldots & \partial_x f^{(2)}(u_N) \\ \partial_y f^{(2)}(u_1) & \partial_y f^{(2)}(u_2) & \ldots & \partial_x f^{(2)}(u_N) \\ \partial_x f^{(3)}(u_1) & \partial_x f^{(3)}(u_2) & \ldots & \partial_x f^{(3)}(u_N)\\ \partial_y f^{(3)}(u_1) & \partial_y f^{(3)}(u_2) & \ldots & \partial_y f^{(3)}(u_N) \end{array} \right] \]

Parameters

[in]	u	gsMatrix of size n x N, where each column of u represents one evaluation point.
[out]	result	gsMatrix of size (n * m) x N. Each row of result corresponds to one component in the target space and contains the gradients for each evaluation point, as row vectors, one after the other (see above for details on the format).

Warning

By default, gsFunction uses central finite differences with h=0.00001! One must override this function in derived classes to get proper results.

Reimplemented from gsFunctionSet< T >.

Reimplemented in gsGeometry< T >, gsFunctionExpr< T >, gsConstantFunction< T >, gsPreCICEFunction< T >, gsAffineFunction< T >, gsMappedSingleSpline< d, T >, gsGeometrySlice< T >, gsSquaredDistance< T >, gsBasisFun< T >, gsGeometryTransform< T >, and gsFuncCoordinate< T >.

virtual short_t domainDim ( ) const

pure virtualinherited

Dimension of the (source) domain.

Returns

For \(f:\mathbb{R}^n\rightarrow\mathbb{R}^m\), returns \(n\).

Implemented in gsTHBSplineBasis< d, T >, gsPatchIdField< T >, gsGeometry< T >, gsParamField< T >, gsNormalField< T >, gsMultiBasis< T >, gsMultiBasis< real_t >, gsMultiPatch< T >, gsMultiPatch< real_t >, gsFsiLoad< T >, gsMaterialMatrixEvalSingle< T, out >, gsFunctionExpr< T >, gsMaterialMatrixIntegrateSingle< T, out >, gsTensorBSplineBasis< 1, T >, gsJacDetField< T >, gsDetFunction< T >, gsConstantFunction< T >, gsRationalBasis< SrcT >, gsRationalBasis< gsBSplineTraits< d, T >::Basis >, gsRationalBasis< gsBSplineBasis< T > >, gsRemapInterface< T >, gsGradientField< T >, gsShellStressFunction< T >, gsPiecewiseFunction< T >, gsLagrangeBasis< T >, gsTensorBasis< d, T >, gsTensorBasis< 1, T >, gsCPPInterface< T >, gsSquaredDistance< T >, gsHBSplineBasis< d, T >, gsAffineFunction< T >, gsPreCICEFunction< T >, gsElementErrorPlotter< T >, gsMappedSingleBasis< d, T >, gsMappedSingleSpline< d, T >, gsConstantBasis< T >, gsSurface< T >, gsLegendreBasis< T >, gsBasisFun< T >, gsCurve< T >, gsAbsError< T >, gsVolume< T >, gsBulk< T >, gsGeometryTransform< T >, gsFuncCoordinate< T >, gsGeometrySlice< T >, gsCauchyStressFunction< T >, and gsMonomialBasis< T >.

gsMatrix< T > eval ( const gsMatrix< T > &  u ) const

inherited

Evaluate the function,.

See Also

eval_into()

virtual void eval_into ( const gsMatrix< T > &  u, gsMatrix< T > &  result  ) const

pure virtual

Evaluate the function at points u into result.

Let n be the dimension of the source space ( n = domainDim() ).
Let m be the dimension of the image/target space ( m = targetDim() ).
Let N denote the number of evaluation points.

Parameters

[in]	u	gsMatrix of size n x N, where each column of u represents one evaluation point.
[out]	result	gsMatrix of size m x N, where each column of u represents the result of the function at the respective valuation point.

Reimplemented from gsFunctionSet< T >.

Implemented in gsPatchIdField< T >, gsMaterialMatrixEvalSingle< T, out >, gsParamField< T >, gsShellStressFunction< T >, gsMaterialMatrixIntegrateSingle< T, out >, gsNormalField< T >, gsFsiLoad< T >, gsFunctionExpr< T >, gsGeometry< T >, gsPiecewiseFunction< T >, gsConstantFunction< T >, gsJacDetField< T >, gsDetFunction< T >, gsRemapInterface< T >, gsGradientField< T >, gsCPPInterface< T >, gsAffineFunction< T >, gsPreCICEFunction< T >, gsMappedSingleSpline< d, T >, gsCauchyStressFunction< T >, gsBasisFun< T >, gsGeometrySlice< T >, gsGeometryTransform< T >, gsFuncCoordinate< T >, gsAbsError< T >, gsElementErrorPlotter< T >, and gsSquaredDistance< T >.

std::vector< gsMatrix< T > > evalAllDers ( const gsMatrix< T > &  u, int  n, bool  sameElement = false  ) const

inherited

Evaluate all derivatives upto order n,.

See Also

evalAllDers_into

void evalAllDers_into ( const gsMatrix< T > &  u, int  n, std::vector< gsMatrix< T > > &  result, bool  sameElement = false  ) const

virtualinherited

Evaluate the nonzero functions and their derivatives up to order n at points u into result.

The derivatives (the 0-th derivative is the function value) are stored in a result. result is a std::vector, where result[i] is a gsMatrix which contains the i-th derivatives.

The entries in result[0], result[1], and result[2] are ordered as in eval_into(), deriv_into(), and deriv2_into(), respectively. For i > 2, the derivatives are stored in lexicographical order, e.g. for order i = 3 and dimension 2 the derivatives are stored as follows: \( \partial_{xxx}, \, \partial_{xxy}, \, \partial_{xyy}, \, \partial_{yyy}.\, \)

Parameters

[in]	u	Evaluation points, each column corresponds to one evaluation point.
[in]	n	All derivatives up to order n are computed and stored in result.
[in,out]	result	See above for format.

Reimplemented in gsTensorBSplineBasis< 1, T >, gsTensorBasis< d, T >, gsTensorBasis< 1, T >, gsGeometry< T >, gsMappedSingleBasis< d, T >, gsConstantFunction< T >, gsTHBSplineBasis< d, T >, gsMappedSingleSpline< d, T >, and gsSquaredDistance< T >.

virtual gsMatrix<T> hessian ( const gsMatrix< T > &  u, index_t  coord = 0  ) const

inlinevirtual

Evaluates the Hessian (matrix of second partial derivatives) of coordinate coord at points u.

void invertPoints ( const gsMatrix< T > &  points, gsMatrix< T > &  result, const T  accuracy = 1e-6, const bool  useInitialPoint = false  ) const

virtual

Takes the physical points and computes the corresponding parameter values. If the point cannot be inverted (eg. is not part of the geometry) the corresponding parameter values will be undefined

gsMatrix< T > laplacian ( const gsMatrix< T > &  u ) const

virtual

Evaluate the Laplacian at points u.

By default uses central finite differences with h=0.00001

Reimplemented in gsFunctionExpr< T >.

int newtonRaphson	(	const gsVector< T > &	value,
		gsVector< T > &	arg,
		bool	withSupport = true,
		const T	accuracy = 1e-6,
		int	max_loop = 100,
		T	damping_factor = 1
	)		const

Newton-Raphson method to find a solution of the equation f(arg) = value with starting vector arg. If the point cannot be inverted the corresponding parameter values will be undefined

void recoverPoints	(	gsMatrix< T > &	xyz,
		gsMatrix< T > &	uv,
		index_t	k,
		const T	accuracy = 1e-6
	)		const

Recovers a point on the (geometry) together with its parameters uv, assuming that the k-th coordinate of the point xyz is not known (and has a random value as input argument).

index_t size ( ) const

inlinevirtual

size

Warning

gsFunction and gsGeometry have size() == 1. This should not be confused with the size eg. of gsGeometry::basis(), which is the number of basis functions in the basis

Returns

the size of the function set: the total number of functions

Reimplemented from gsFunctionSet< T >.

Reimplemented in gsPiecewiseFunction< T >.

virtual short_t targetDim ( ) const

inlinevirtualinherited

Dimension of the target space.

Returns

For \(f:\mathbb{R}^n\rightarrow\mathbb{R}^m\), returns \(m\).

Reimplemented in gsPatchIdField< T >, gsGeometry< T >, gsParamField< T >, gsNormalField< T >, gsMultiBasis< T >, gsMultiBasis< real_t >, gsMultiPatch< T >, gsMultiPatch< real_t >, gsFsiLoad< T >, gsMaterialMatrixEvalSingle< T, out >, gsMaterialMatrixIntegrateSingle< T, out >, gsFunctionExpr< T >, gsJacDetField< T >, gsDetFunction< T >, gsConstantFunction< T >, gsShellStressFunction< T >, gsGradientField< T >, gsPiecewiseFunction< T >, gsSquaredDistance< T >, gsAffineFunction< T >, gsPreCICEFunction< T >, gsMappedSingleSpline< d, T >, gsBasisFun< T >, gsAbsError< T >, gsGeometrySlice< T >, gsGeometryTransform< T >, gsFuncCoordinate< T >, and gsCauchyStressFunction< T >.