Re: [ublas] vector of vector adaptors

From: Max Weinberg (maximilian.weinberg_at_[hidden])
Date: 2007-07-18 07:40:34

There are some missing features with gvov. When I last looked at it, which is 
already quite a while ago: There was no push_back, which would enable fast 
sequential construction of sparse matrices. If the outer vector was sparse, it 
was ignored, so the outer dimension is handled like a full vector (a similar 
bug is present for the other sparse matrix types). Transpose did not work for 
gvov.
I apologize if any of that is not true anymore. In general, gvov can easily be 
the fastest of the sparse matrix types, if access is done primarily along one 
dimension.
Greetings, 
Max