Hi everybody !!

I am new using eigen, and I have a mex function where i obtain a sparse matrix from matlab
and i want to calculate the LU decomposition (just to try). But I get the following error:

Invalid ordered insertion (invalid inner index)

Code: Select all

   MappedSparseMatrix<double,ColMajor> mat (mxGetM(in[0]), mxGetM(in[0]), mxGetNzmax(in[0]), (int*)mxGetJc(in[0]), (int*)mxGetIr(in[0]), mxGetPr(in[0]));

   VectorXd bb, xx;
   // fill b
   // solve Ax = b
   SparseLU< MappedSparseMatrix<double,ColMajor>> solver;
   solver.compute(mat);


The problem comes in the last line.

Thanks in advance

Which Eigen version? Maybe the entries of mat are not sorted? You can sort them by copying it to a SparseMatrix<double,RowMajor>.

I use the last version and i get the pointers from the mex function obtained by matlab. So the matrix is sorted according to matlab.This is the reason why i use the mapped version of the sparsematrix class. So how could i sort it out if i create the matrix with the mex pointers. Or how could i sort it out if it is the mapped sparse class.

Thank you very much for your help

I would prefer not to do it by copying point by point in another matrix because it is a huge matrix and i would like to have only one matrix.

Thank you very much

Is it really guaranteed that the entries are sorted? Moreover, Eigen is 0-index based, while MatLab is 1-index based, so you have to adjust the index buffers (decrement them, do the work, increment them back).

Hi,

thank you very much for your answer. I think i have a problem of concept so i made a short example to see what i am doing wrong

Code: Select all

 MappedSparseMatrix<double,ColMajor> mat (mxGetM(in[0]), mxGetM(in[0]), mxGetNzmax(in[0]), (int*)yy, (int*)xx, (double*)mxGetPr(in[0]));


yy: are the indexs to rows
xx are the indexs to columns

matrix is
[2,1
4,1]

and xx
xx[0] 0
xx[1] 1
xx[2] 0
xx[3] 1
and yy
yy[0] 0
yy[1] 0
yy[2] 1
yy[3] 1

so it is column major. Am i doing something wrong?.

Thanks again

the yy index vector does not correspond to what Eigen expects. In your example it should be :

yy[] = [0,2,4]

see http://eigen.tuxfamily.org/dox/group__T ... parse.html, "compressed" exemple.

Thanks for your answer. I also tried like this but it seems this function has a bug because there is no way to make it work even for simple
examples

Code: Select all

 MappedSparseMatrix<double,ColMajor> mat (mxGetM(in[0]), mxGetM(in[0]), mxGetNzmax(in[0]), (int*)yy, (int*)xx, (double*)L);


where
yy[0] 0
yy[1] 2
yy[2] 4
xx[0] 0
xx[1] 1
xx[2] 2
xx[3] 3
L[0] 2
L[1] 4
L[2] 1
L[3] 1

even this small example does not work. It does not get the indexes correctly. I cannot even read it

Code: Select all

         for (int k=0; k<mat.outerSize(); ++k){

   for (MappedSparseMatrix<double>::InnerIterator it(mat,k); it; ++it)
   {
   int a = it.value();
   a = it.row(); // row index
   a = it.col(); // col index (here it is equal to k)
   a = it.index(); // inner index, here it is equal to it.row()
   }
   }


This small code crashes.

Thanks again

What are mat.rows() and mat.cols() ? If 2x2 then xx should be 0,1,0,1.

Thank you very much. Yeah it is right I was writing wrong the indices and I also had an error casting of the pointers !!!. With this I solve my problem.
Thank you so much for helping me understand the indexes to the sparse matrix. It works excellent now for my problem !!!!

See you !!!