Hello, I am no sure about these two issues:

1) I have a symmetric sparse matrix (double) M in which I store only the upper part and a non symmetric sparse matrix (double) T ;
Is T.transpose()*M.selfadjointView<Upper>()*T correct ?

2) In the case of complex matrix, to use selfadjointView<Symmetrix > in place of selfadjointView<Upper> is equivalent to force:

mij = mji = a+ib ?

Thx

1) T.transpose()*M.selfadjointView<Upper>()*T probably does not work yet. You have to evaluate M.selfadjointView<Upper>() into a full SparseMatrix:

SparseMatrix<double> tmp;
tmp = M.selfadjointView<Upper>();
T.transpose()*tmp*T


2) selfadjointView<...> only accepts Upper or Lower. In the future, symmetricView<Upper or Lower> will provide what you're looking for.

ggael wrote:1) T.transpose()*M.selfadjointView<Upper>()*T probably does not work yet. You have to evaluate M.selfadjointView<Upper>() into a full SparseMatrix:

SparseMatrix<double> tmp;
tmp = M.selfadjointView<Upper>();
T.transpose()*tmp*T


2) selfadjointView<...> only accepts Upper or Lower. In the future, symmetricView<Upper or Lower> will provide what you're looking for.

tnx