Is it possible to specify the Scalar template parameter of a SparseMatrix as an Eigen::Matrix?

For example, I have the following code I'm trying to get to work.

----------------------------------------------------
#include <Eigen/Dense>
#include <Eigen/SparseCore>
#include <iostream>

using namespace Eigen;
using namespace std;
int main(){

Matrix<double,3,3> dmat;
SparseMatrix<Matrix<double,3,3> > smat(2,2);

dmat.fill(1.0);

smat.insert(0,0) = dmat;

return 0;
}
----------------------------------------------------

And I get the following error:

Matrix.h: In member function ‘typename Eigen::SparseMatrix<_Scalar, _Flags, _Index>::Scalar& Eigen::SparseMatrix<_Scalar, _Flags, _Index>::insertUncompressed(typename Eigen::internal::traits<Eigen::SparseMatrix<_Scalar, _Options, _Index> >::Index, typename Eigen::internal::traits<Eigen::SparseMatrix<_Scalar, _Options, _Index> >::Index) [with _Scalar = Eigen::Matrix<double, 3, 3, 0, 3, 3>, int _Options = 0, _Index = int]’:
/home/flyntbt/Desktop/Other/eigen-eigen-ffa86ffb5570/Eigen/src/SparseCore/SparseMatrix.h:228: instantiated from ‘typename Eigen::internal::traits<Eigen::SparseMatrix<_Scalar, _Options, _Index> >::Scalar& Eigen::SparseMatrix<_Scalar, _Flags, _Index>::insert(typename Eigen::internal::traits<Eigen::SparseMatrix<_Scalar, _Options, _Index> >::Index, typename Eigen::internal::traits<Eigen::SparseMatrix<_Scalar, _Options, _Index> >::Index) [with _Scalar = Eigen::Matrix<double, 3, 3, 0, 3, 3>, int _Options = 0, _Index = int]’
test.cpp:42: instantiated from here
/home/flyntbt/Desktop/Other/eigen-eigen-ffa86ffb5570/Eigen/src/SparseCore/SparseMatrix.h:1146: error: no match for ‘operator=’ in ‘((Eigen::SparseMatrix<Eigen::Matrix<double, 3, 3, 0, 3, 3>, 0, int>*)this)->Eigen::SparseMatrix<Eigen::Matrix<double, 3, 3, 0, 3, 3>, 0, int>::m_data.Eigen::internal::CompressedStorage<_Scalar, _Index>::value [with _Scalar = Eigen::Matrix<double, 3, 3, 0, 3, 3>, _Index = int](((long unsigned int)p)) = 0’
/home/flyntbt/Desktop/Other/eigen-eigen-ffa86ffb5570/Eigen/src/Core/Matrix.h:154: note: candidates are: Eigen::Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols>& Eigen::Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols>::operator=(const Eigen::Matrix<_Scalar, _Rows, _Cols, _Options, _MaxRows, _MaxCols>&) [with _Scalar = double, int _Rows = 3, int _Cols = 3, int _Options = 0, int _MaxRows = 3, int _MaxCols = 3]

Any Ideas how to make this work?

Thanks.

Look at the BlockSparseMatrix class in unsupported/SparseExtra module of the devel branch: http://eigen.tuxfamily.org/dox-devel/un ... atrix.html

Using this class will be much more powerful since it is compatible with products and iterative solvers.

btw, using an Array instead of a Matrix should fix your issue: SparseMatrix<Array<double,3,3> >. However, the semantic of operator* will be different.

Thanks for the help.
I downloaded the latest dev version to get the BlockSparseMatrix and experimented with it. It is a long way from being ready since many of the basic functions (insert, add, matrix product, etc.) do not work. I also tried the SparseMatrix<Array...>> possibility but was really looking for matrix * matrix functionality.

A question:
What is the problem with adding the ability to use a Matrix as the Scalar parameter to a SparseMatrix? There must be some serious problem since an Array works well yet the developers still chose to create a whole new data type to handle the Matrix situation. Why not just provide some template specializations to handle the cases when the Scalar is a Matrix and the default methods are not what is needed?