I think this is a silly question, but is there a way to read-access sparse matricies where I can do

mySparse.someFunc(row,col)
and if there is a coeff, it will return the value, and if its empty, it will return a 0?

I ask because nothing here (http://eigen.tuxfamily.org/dox-devel/cl ... mbers.html) appears to solve this problem

To provide a little background, I have some lookup tables that are 20-40% sparse and I wanted to index into them for the interpolation functions in a reasonably logical way and in this case we would benefit by saving the memory at the expense of extra cycles on the lookup.

thanks

In order to keep things in one thread, my current (sub-optimal) solution crashes the vs2010 compiler itself.

My current solution is to allocate a dense vector, setZero, then insert the non-zero coeffs from the sparse into the dense.

I'm trying to compile my code using VS2010 x64 target with optimization off and the compiler itself crashes with this error.

1>c:\users\USERNAME\desktop\local_code\branches\onlinecarving\thirdparty\eigen\eigen\src/Sparse/SparseTranspose.h(48): fatal error C1001: An internal error has occurred in the compiler.
1> (compiler file 'msc1.cpp', line 1420)
1> To work around this problem, try simplifying or changing the program near the locations listed above.
1> Please choose the Technical Support command on the Visual C++
1> Help menu, or open the Technical Support help file for more information
1> c:\users\USERNAME\desktop\local_code\branches\onlinecarving\thirdparty\eigen\eigen\src/Sparse/SparseTranspose.h(53) : see reference to class template instantiation 'Eigen::TransposeImpl<MatrixType,Eigen::Sparse>::InnerIterator' being compiled
========== Build: 0 succeeded, 1 failed, 0 up-to-date, 0 skipped ==========

Something like....

Code: Select all

std::vector< Eigen::SparseMatrix<float, RowMajor> >      m_lut;
....init stuff....
Eigen::VectorXf oneRow(m_lut[k].innerSize());
oneRow.setZero();
for (int j=0; j<m_lut[k].outerSize(); ++j)
{
   for (SparseMatrixType::InnerIterator it(m_lut[k],j); it; ++it)
   {
      oneRow(it.col()) = it.value();
   }
}

..........
Thoughts?

for the first question there is a .coeff(i,j) method that does exactly that.

ggael wrote:for the first question there is a .coeff(i,j) method that does exactly that.

Well I feel stupid. Shouldn't that show up on that documentation page, or is my link above to a page which doesn't list all the members?