[ublas] generic (sparse) index type

From: Gunter Winkler (guwi17_at_[hidden])
Date: 2006-11-28 06:40:45

Hallo,

using non standard types as size_type (or index type) in sparse matrices 
implies using non standard types in row_major. Thus the correct declaration 
of a compressed matrix should look like:

typedef compressed_matrix<
    double, 
    basic_row_major<unsigned INT, signed INT>,
    0,
    unbounded_array<unsigned INT>,
    unbounded_array<double>
    >  SPARSE;

where INT is any integer type. Note that I replaced the default row_major 
(which is basic_row_major<size_t, ptr_diff_t>) by basic_row_major<Z,D>. 

This change breaks some specializations of axpy_prod which use row_major and 
column_major instead of the generic basic_row_major<Z,D> and 
basic_column_major<Z,D>. 
(Z ... integral type: 0 <= z <= max, D ... integral type: -max <= d <= max)

Similar changes should be made for each occurrence of row_major/column_major. 

mfg
Gunter