Introduce an unpack to arguments operation for vectors and matrices

[nasos_i@…]

The new ublas assignment facility provides a convenient way to fill up matrices and vectors, but the opposite would also be useful. Quoting:

The boost assignment is really nice, but what about the other direction
(from vector to scalars)? This comes up quite often.

ublas::vector<double> f()
{
ublas::vector<double> v(2);
 v(0) = 1.1;
v(1) = 2.1;
return v;
}

double x;
double y;

 ublas::vector<double> v(2);
v(0) = 1.1;
v(1) = 2.1;

ublas::tie(x, y) = v;
ublas::tie(x, y) = f(); //With function

ublas::matrix<double> mat(2,2); //....
ublas::tie(x, y) = row(mat, 0);

ublas::vector<int> v2(2);
v2(0) = 1;
 v2(1) = 2;

        int z;
        ublas::tie(x, z); //With different types

For vector expressions, this seems to be easy enough to do. See the
attached basic implementation for 2 parameters. With a little boost
preprocessor, this could be fairly general. What do people think? Are
there any boost PP who could generalize this?

-Jesse 

and

What's about ublas::tie(V1,V2,V3) = M;
That could be useful for applications where Vn are not directly managed by
ublas. I mean that you make matrix computations which vectors are then sent
to some sort of external outputs. That would be written in one line (like
controlling and engine or whatever you can imagine, you know).
that's splitting up a matrix in vectors .... ?

Cheers,
David 

Some ideas of implementation tools and techniques include:

1. Boost tuple
2. Variadic templates
3. Custom exhaustive implementation (up to a certain number of return values)