dolfin team mailing list archive

["Garth N. Wells" <gnw20@xxxxxxxxx>]

Bartosz Sawicki wrote:
> Hi there,
>
> I've found that VTK export in Dolfin has one serious limitation. It 
> ignores type of finite element of DiscreteFunction and always 
> interpolates function into vertex values.
> That is very general solution, which always produce something which can 
> be visualized, but in some cases (eg. Lagrange element - 0 order) such 
> procedure induces errors.
> VTK file format specification allows two types of functions PointData 
> and CellData. Paraview also visualize correctly both types. So why do we 
> alway transform everything to PointData ?
> I've already write some code that enable this feature in Dolfin. 
> Unfortunatelly it was not easy, some special hacks was needed. Of course 
> I can send path or bundle, but I this can be done better after some 
> redesign.
> The main problem is that, when I have object of DiscreteFunction I can't 
> check what kind of finite element is used inside, or maybe I omit 
> something?  So I can't write code which will automatically detect when 
> to use PointData, or CellData. At this moment I solve this problem by 
> introducing new File::Type (vtkc). When new File object is created user 
> has possibility to choose whether it is normal VTK with PointData, or 
> VTK with CellData.
> What do you think about such solution?
I don't think that this is the best solution. Ideally, a Function will 
know what's appropriate. We previously supported cell data (Kristian 
implemented it) but we removed it because we couldn't determine robustly 
for a general Function (possibly a mixed element) whether cell or point 
data was appropriate. This may now be easier now as a lot of changes 
have been made since we removed support for cell data.
Once a Function can determine the appropriate format, formatting the 
output is pretty straightforward. For now, just project solutions onto a 
continuous P1 basis.
Garth



> regrd.
> BArtosz
> _______________________________________________
> DOLFIN-dev mailing list
> DOLFIN-dev@xxxxxxxxxx
> http://www.fenics.org/mailman/listinfo/dolfin-dev