A trampoline class for mfem::Matrix. More...

#include <PyMatrix.h>

Inheritance diagram for serif::pybind::PyMatrix:

Public Member Functions
mfem::real_t &	Elem (int i, int j) override
	Access element (i,j) for read/write. Pure virtual in mfem::Matrix. Must be overridden in Python.

const mfem::real_t &	Elem (int i, int j) const override
	Access element (i,j) for read-only. Pure virtual in mfem::Matrix. Must be overridden in Python.

mfem::MatrixInverse *	Inverse () const override
	Get the inverse of the matrix. Pure virtual in mfem::Matrix. Must be overridden in Python. The caller is responsible for deleting the returned MatrixInverse object.

void	Finalize (int skip_zeros) override
	Finalize matrix assembly. For sparse matrices, this typically involves finalizing the sparse structure. Can be overridden in Python.

void	Mult (const mfem::Vector &x, mfem::Vector &y) const override
	Perform the operator action: y = A*x. Inherited from mfem::Operator. Can be overridden in Python. If not overridden, mfem::Matrix's default implementation is used.

void	MultTranspose (const mfem::Vector &x, mfem::Vector &y) const override
	Perform the transpose operator action: y = A^T*x. Inherited from mfem::Operator. Can be overridden in Python.

void	AddMult (const mfem::Vector &x, mfem::Vector &y, const mfem::real_t a=1.0) const override
	Perform the action y += a(Ax). Inherited from mfem::Operator. Can be overridden in Python.

void	AddMultTranspose (const mfem::Vector &x, mfem::Vector &y, const mfem::real_t a=1.0) const override
	Perform the action y += a(A^Tx). Inherited from mfem::Operator. Can be overridden in Python.

mfem::Operator &	GetGradient (const mfem::Vector &x) const override
	Get the gradient operator (Jacobian) at a given point x. Inherited from mfem::Operator. Can be overridden in Python. For a linear matrix operator, the gradient is typically the matrix itself.

void	AssembleDiagonal (mfem::Vector &diag) const override
	Assemble the diagonal of the operator. Inherited from mfem::Operator. Can be overridden in Python.

const mfem::Operator *	GetProlongation () const override
	Get the prolongation operator. Inherited from mfem::Operator. Can be overridden in Python.

const mfem::Operator *	GetRestriction () const override
	Get the restriction operator. Inherited from mfem::Operator. Can be overridden in Python.

void	RecoverFEMSolution (const mfem::Vector &X, const mfem::Vector &b, mfem::Vector &x) override
	Recover the FEM solution. Inherited from mfem::Operator. Can be overridden in Python.

Detailed Description

A trampoline class for mfem::Matrix.

This class allows Python classes to inherit from mfem::Matrix and correctly override its virtual functions, including those inherited from its base, mfem::Operator. This is useful for creating custom matrix types in Python that can interact seamlessly with MFEM's C++ components.

See also: mfem::Matrix; mfem::Operator; PyOperator

Python Usage Example:: import mfem.ser_ext as mfem

import numpy as np

class MyPythonMatrix(mfem.Matrix):

def __init__(self, size=0):

super().__init__(size) # Call base C++ constructor

if size > 0:

# Example: Initialize with some data if size is provided

# Note: Direct data manipulation might be complex due to

# C++ ownership. This example is conceptual.

# For a real-world scenario, you might manage data in Python

# and use Eval or Mult to reflect its state.

self._py_data = np.zeros((size, size))

else:

self._py_data = None

# --- mfem.Matrix pure virtual overrides ---

def Elem(self, i, j):

# This is problematic for direct override for write access

# due to returning a reference.

# Typically, you'd manage data and use it in Mult/GetRow etc.

# For read access (const version), it's more straightforward.

if self._py_data is not None:

return self._py_data[i, j]

raise IndexError("Matrix data not initialized or index out of bounds")

# const Elem version for read access

# def GetElem(self, i, j): # A helper in Python if Elem is tricky

# return self._py_data[i,j]

def Inverse(self):

# Return an mfem.MatrixInverse object

# This is a simplified example; a real inverse is complex.

print("MyPythonMatrix.Inverse() called, returning dummy inverse.")

# For a real implementation, you'd compute and return an actual

# mfem.MatrixInverse or a Python-derived version of it.

# This might involve creating a new PyMatrix representing the inverse.

identity_inv = mfem.DenseMatrix(self.Width())

for i in range(self.Width()):

identity_inv[i,i] = 1.0

return mfem.MatrixInverse(identity_inv) # Example

# --- mfem.Matrix regular virtual overrides ---

def Finalize(self, skip_zeros=1):

super().Finalize(skip_zeros) # Call base

print(f"MyPythonMatrix.Finalize({skip_zeros}) called.")

# --- mfem.Operator virtual overrides ---

def Mult(self, x, y):

# x is mfem.Vector (input), y is mfem.Vector (output)

if self._py_data is not None:

x_np = x.GetDataArray()

y_np = np.dot(self._py_data, x_np)

y.SetSize(self._py_data.shape[0])

y.Assign(y_np)

else:

# Fallback to base class if no Python data

# super().Mult(x,y) # This might not work as expected if base is abstract

# or if the C++ mfem.Matrix itself doesn't have data.

# For a sparse matrix, this would call its sparse Mult.

# For a dense matrix, it would use its data.

# If this PyMatrix is purely Python defined, it must implement Mult.

raise NotImplementedError("Mult not implemented or data not set")

print("MyPythonMatrix.Mult called.")

# Using the Python Matrix

mat_size = 3

py_mat = MyPythonMatrix(mat_size)

py_mat._py_data = np.array([[1,2,3],[4,5,6],[7,8,9]], dtype=float)

x_vec = mfem.Vector(mat_size)

x_vec.Assign(np.array([1,1,1], dtype=float))

y_vec = mfem.Vector(mat_size)

py_mat.Mult(x_vec, y_vec)

print("Result y:", y_vec.GetDataArray())

# inv_op = py_mat.Inverse() # Be cautious with dummy implementations

# print("Inverse type:", type(inv_op))

# Note: Overriding Elem(i,j) to return a C++ reference from Python

# for write access (mfem::real_t&) is complex and often not directly feasible

# in a straightforward way with pybind11 for typical Python data structures.

# Python users would typically interact via methods like Set, Add, or by

# providing data that the C++ side can access (e.g., via GetData).

# The const version of Elem (read-only) is more manageable.

Definition at line 118 of file PyMatrix.h.

Member Function Documentation

◆ AddMult()

void serif::pybind::PyMatrix::AddMult	(	const mfem::Vector &	x,
		mfem::Vector &	y,
		const mfem::real_t	a = 1.0 ) const

override

Perform the action y += a*(A*x). Inherited from mfem::Operator. Can be overridden in Python.

Parameters

x	The input vector.
y	The vector to which a(Ax) is added.
a	Scalar multiplier (defaults to 1.0).

Note: PYBIND11_OVERRIDE is used in the .cpp file.

Definition at line 70 of file PyMatrix.cpp.

◆ AddMultTranspose()

void serif::pybind::PyMatrix::AddMultTranspose	(	const mfem::Vector &	x,
		mfem::Vector &	y,
		const mfem::real_t	a = 1.0 ) const

override

Perform the action y += a*(A^T*x). Inherited from mfem::Operator. Can be overridden in Python.

Parameters

x	The input vector.
y	The vector to which a(A^Tx) is added.
a	Scalar multiplier (defaults to 1.0).

Note: PYBIND11_OVERRIDE is used in the .cpp file.

Definition at line 81 of file PyMatrix.cpp.

◆ AssembleDiagonal()

void serif::pybind::PyMatrix::AssembleDiagonal ( mfem::Vector & diag ) const

override

Assemble the diagonal of the operator. Inherited from mfem::Operator. Can be overridden in Python.

Parameters

diag	Output vector to store the diagonal entries.

Note: PYBIND11_OVERRIDE is used in the .cpp file.

Definition at line 101 of file PyMatrix.cpp.

◆ Elem() [1/2]

const mfem::real_t & serif::pybind::PyMatrix::Elem	(	int	i,
		int	j ) const

override

Access element (i,j) for read-only. Pure virtual in mfem::Matrix. Must be overridden in Python.

Parameters

i	Row index.
j	Column index.

Returns: Const reference to the matrix element.

Note: PYBIND11_OVERRIDE_PURE is used in the .cpp file.

Definition at line 18 of file PyMatrix.cpp.

◆ Elem() [2/2]

mfem::real_t & serif::pybind::PyMatrix::Elem	(	int	i,
		int	j )

override

Access element (i,j) for read/write. Pure virtual in mfem::Matrix. Must be overridden in Python.

Parameters

i	Row index.
j	Column index.

Returns: Reference to the matrix element.

Note: Returning a C++ reference from Python for write access can be complex. Consider alternative data management strategies in Python. PYBIND11_OVERRIDE_PURE is used in the .cpp file.

Definition at line 8 of file PyMatrix.cpp.

◆ Finalize()

void serif::pybind::PyMatrix::Finalize ( int skip_zeros )

override

Finalize matrix assembly. For sparse matrices, this typically involves finalizing the sparse structure. Can be overridden in Python.

Parameters

skip_zeros See mfem::SparseMatrix::Finalize documentation.

Note: PYBIND11_OVERRIDE is used in the .cpp file.

Definition at line 38 of file PyMatrix.cpp.

◆ GetGradient()

mfem::Operator & serif::pybind::PyMatrix::GetGradient ( const mfem::Vector & x ) const

override

Get the gradient operator (Jacobian) at a given point x. Inherited from mfem::Operator. Can be overridden in Python. For a linear matrix operator, the gradient is typically the matrix itself.

Parameters

x	The point at which to evaluate the gradient (often unused for linear operators).

Returns: A reference to the gradient operator.

Note: PYBIND11_OVERRIDE is used in the .cpp file.

Definition at line 92 of file PyMatrix.cpp.

◆ GetProlongation()

const mfem::Operator * serif::pybind::PyMatrix::GetProlongation ( ) const

override

Get the prolongation operator. Inherited from mfem::Operator. Can be overridden in Python.

Returns: A const pointer to the prolongation operator, or nullptr if not applicable.

Note: PYBIND11_OVERRIDE is used in the .cpp file.

Definition at line 110 of file PyMatrix.cpp.

◆ GetRestriction()

const mfem::Operator * serif::pybind::PyMatrix::GetRestriction ( ) const

override

Get the restriction operator. Inherited from mfem::Operator. Can be overridden in Python.

Returns: A const pointer to the restriction operator, or nullptr if not applicable.

Note: PYBIND11_OVERRIDE is used in the .cpp file.

Definition at line 118 of file PyMatrix.cpp.

◆ Inverse()

mfem::MatrixInverse * serif::pybind::PyMatrix::Inverse ( ) const

override

Get the inverse of the matrix. Pure virtual in mfem::Matrix. Must be overridden in Python. The caller is responsible for deleting the returned MatrixInverse object.

Returns: Pointer to an mfem::MatrixInverse object.

Note: PYBIND11_OVERRIDE_PURE is used in the .cpp file.

Definition at line 28 of file PyMatrix.cpp.

◆ Mult()

void serif::pybind::PyMatrix::Mult	(	const mfem::Vector &	x,
		mfem::Vector &	y ) const

override

Perform the operator action: y = A*x. Inherited from mfem::Operator. Can be overridden in Python. If not overridden, mfem::Matrix's default implementation is used.

Parameters

x	The input vector.
y	The output vector (result of A*x).

Note: PYBIND11_OVERRIDE is used in the .cpp file.

Definition at line 49 of file PyMatrix.cpp.

◆ MultTranspose()

void serif::pybind::PyMatrix::MultTranspose	(	const mfem::Vector &	x,
		mfem::Vector &	y ) const

override

Perform the transpose operator action: y = A^T*x. Inherited from mfem::Operator. Can be overridden in Python.

Parameters

x	The input vector.
y	The output vector (result of A^T*x).

Note: PYBIND11_OVERRIDE is used in the .cpp file.

Definition at line 60 of file PyMatrix.cpp.

◆ RecoverFEMSolution()

void serif::pybind::PyMatrix::RecoverFEMSolution	(	const mfem::Vector &	X,
		const mfem::Vector &	b,
		mfem::Vector &	x )

override

Recover the FEM solution. Inherited from mfem::Operator. Can be overridden in Python.

Parameters

X	The reduced solution vector.
b	The right-hand side vector.
x	Output vector for the full FEM solution.

Note: PYBIND11_OVERRIDE is used in the .cpp file.

Definition at line 126 of file PyMatrix.cpp.

The documentation for this class was generated from the following files:

src/python/mfem/Trampoline/PyMFEMTrampolines/Operator/Matrix/PyMatrix.h
src/python/mfem/Trampoline/PyMFEMTrampolines/Operator/Matrix/PyMatrix.cpp

Public Member Functions

Detailed Description

Member Function Documentation

◆ AddMult()

◆ AddMultTranspose()

◆ AssembleDiagonal()

◆ Elem() [1/2]

◆ Elem() [2/2]

◆ Finalize()

◆ GetGradient()

◆ GetProlongation()

◆ GetRestriction()

◆ Inverse()

◆ Mult()

◆ MultTranspose()

◆ RecoverFEMSolution()