dolfin team mailing list archive

[Anders Logg <logg@xxxxxxxxx>]

On Thu, Jun 12, 2008 at 05:41:36PM +0100, Nuno David Lopes wrote:
> Isn't the code:
> ......................
> Matrix A;
> Vector b,x;
> assemble(A,a,mesh);
> assemble(b,L,mesh);
> solver.solve(A,x,b);
> Function up(mesh,x,a); //up=(u,p) velocities and pressure
> Function u=up[0];
> Function p=up[1]; 
> ........................... 
> supposed to work?
> this gives an error:
> ----------------------
> conversion from 'dolfin::SubFunction' to non-scalar type 'dolfin::Function' 
> requested
> --------------------
> 
> On the other hand I know that this works:
> Function u(up[0]);
> Function p(up[1]);

It's the "explicit" keyword in Function.h that prevents this. Without
this, C++ does some magic (like generating assignment operators and
similar). We want to prevent this to be able to control what happens
during assignment. Another reason is to prevent automatic casting
in other cases. See here for example:

  http://www.sjbrown.co.uk/?article=explicit

Martin may have some more comments.

-- 
Anders

Attachment: signature.asc
Description: Digital signature