← Back to team overview

dolfin team mailing list archive

Re: Abstraction idea

 

On Thu, Jun 05, 2008 at 01:06:53PM +0200, Martin Sandve Alnæs wrote:
> It's a bit cumbersome to pass around mesh, dofmap, and form to be able
> to create functions in finite element spaces. Maybe a FunctionSpace
> class could collect the necessary info?
> 
> # Instead of:
> uvec = Vector()
> u = cpp_Function(mesh, uvec, dofmap, form, 0)
> vvec = Vector()
> v = cpp_Function(mesh, vvec, dofmap, form, 0)
> 
> # It would be nice to write:
> U = FunctionSpace(mesh, dofmapset, form, 0) # or something similar
> u = Function(U)
> v = Function(U)

I think it sounds really good, but shouldn't it really be

  V = FunctionSpace(mesh, finite_element, dof_map)

So

  dofmapset --> dof_map
  form --> finite_element

> This concept is probably usable in C++ as well.

Yes, it should be implemented there.

> (Note that I'm not volunteering to implement this myself now, don't
> have the time).

ok. Could you put it on the todo list? Either we can keep editing the
TODO file in DOLFIN, or we can have a TODO list on the wiki. I think
Garth suggested this yesterday.

-- 
Anders


Follow ups

References