[mlpack-svn] [MLPACK] #159: save() and load() for SpMat

[MLPACK Trac]

#159: save() and load() for SpMat
---------------------------------------------+------------------------------
  Reporter:  rcurtin                         |        Owner:  mamidon 
      Type:  enhancement                     |       Status:  accepted
  Priority:  major                           |    Milestone:          
 Component:  armadillo sparse                |   Resolution:          
  Keywords:  sparse matrix save format load  |     Blocking:          
Blocked By:                                  |  
---------------------------------------------+------------------------------

Comment (by rcurtin):

 I'm curious about a change in save_pgm_binary().  PGM is a dense format,
 so we end up writing even the zeros out into a non-sparse format.
 However, when we actually write the data, we do this
 (diskio_meat.hpp:2251):

 {{{
 f.write(reinterpret_cast<const char*>(tmp.mem),
 std::streamsize(x.n_rows*x.n_cols) );
 f.write( reinterpret_cast<const char*>(x.row_indices),
 std::streamsize(x.n_nonzero*sizeof(uword)) );
 f.write( reinterpret_cast<const char*>(x.col_ptrs),
 std::streamsize((x.n_cols+1)*sizeof(uword)) );
 }}}

 Shouldn't the column pointers and row indices be ignored since we're
 saving it densely?

-- 
Ticket URL: <http://trac.research.cc.gatech.edu/fastlab/ticket/159#comment:5>
MLPACK <www.fast-lab.org>
MLPACK is an intuitive, fast, and scalable C++ machine learning library developed by the FASTLAB at Georgia Tech under Dr. Alex Gray.