RE: dolfin function to octave function

[marco Vassallo]

> From: address@hidden
> Subject: dolfin function to octave function
> Date: Thu, 12 Sep 2013 13:29:10 +0200
> To: address@hidden
> CC: address@hidden
>
> Marco,
>
> The following code in the "AD_time" example
>
> ---------------------------------------------------------------
> U = lsode (@(u, t) f(u, t, A, ML), u, time);
>
> for ii = 1:1:numel (time)
> name = sprintf ("u_%3.3d", ii);
> delete ([name ".vtu"]);
> fpl_vtk_write_field (name, msho, {U(ii,:)', 'u'}, {}, 1);
> endfor
> ---------------------------------------------------------------
>
> is a very dirty hack of extremely limited use, it only happens to work
> because the function space you are using for this problem has lagrangian
> degrees of freedom located at the vertices of the tetrahedra.

Hi,

I do it only to use the pkg fpl.
This should work

U = lsode (@(u, t) f(u, t, A, ML), u, time);

for ii = 1:1:numel (time)
 name = sprintf ("u_%3.3d", ii);
 delete ([name ".vtu"]);
 u = Function ('u', V, U(ii,:));
 save (name, u);
endfor

>
> It will fail for higher order elements that have more degrees of freedom per element,
> and it will also fail for elements where the dofs are not collocated at the element vertices,
> like e.g. raviart-thomas.
>
> What is really needed is a method to evaluate the problem solution at a set of points
> defined by their coordinates.
>
> you should add this to the todo list for project completion.
>
> c.