dolfin team mailing list archive

[Dag Lindbo <dag@xxxxxxxxxx>]

Ola Skavhaug wrote:
> Dag Lindbo skrev den 09/04-2008 følgende:
> > Dag Lindbo wrote:
> > > Ola Skavhaug wrote:
> > > > Dag Lindbo skrev den 09/04-2008 følgende:
> > > > > Hello again!
> > > > >
> > > > > Why is it not possible any longer to get the Vector associated with 
> > > > > a Function? E.g:
> > > > >
> > > > > #include <dolfin.h>
> > > > > using namespace dolfin;
> > > > >
> > > > > int main()
> > > > > {
> > > > >    Function u;
> > > > >    Vector& v = u.vector();
> > > > > }
> > > > int main()
> > > > {
> > > >      Function u;
> > > >      Vector v;
> > > >      v = u.vector();
> > > >      return 0;
> > > > }
> > > >
> > > > Should work. A Function returns a GenericVector reference, and the Vector
> > > > reference needs to be a Vector. Then the operator= in Vector is  
> > > > applied in the
> > > > assignment.
> > > >
> > > > Ola
> > > Ola,
> > >
> > > Sorry if I'm this is a stupid/incorrect observation. What you propose 
> > > is a a deep copy of the vector. So something like
> > >
> > > uBlasVector& vv = *(v.instance());
> > > vv[0] = 1000.32;
> > >
> > > has no impact on the data owned by u. uBlas only has deep copy  
> > > semantics, which are invoked in dolfin::Vector::operator= (i.e. *vector 
> > > = *x; )
> > >
> > > Is this really what we want?
> > >
> > > /Dag
> > Seems I'm also sloppy with grammar today. My previous post may sound  
> > offensive towards Ola, but I only meant to pre-excuse _myself_ in case  
> > my understanding of C++ has deserted me this evening. /Dag
>
> No offence taken :) My point was simply that since the underlying
> GenericVector is not a Vector, the lines:
>
> Function u;
> Vector& v = u.vector();
>
> does not work. But the compiler told you that already. Do you want a method
> for rebinding the DefaultVector* vector attribute of Vector? That way, the
> deep copy will be avoided.
>
> Ola

No, I simply want to retrieve the Vector reference associated with a 
DiscreteFunction, so as to retain the reference semantics that used to 
be present in the LA interface. I realize now that it had to be changed 
out of necessity. For my purposes I now do

Vector& x = dynamic_cast<Vector&>(u.vector());
instead of
Vector& x = u.vector();

and

uBlasVector& xx = *x.instance();
instead of
uBlasVector& xx = x.vec();

Would anyone advise against this? I'm happy that it relieves me of 
dealing with explicit pointers (only one dereference), and "hides" the 
GenericVector.

Thanks a lot for your input guys!

/Dag