Plots of mesh functions over facets look wrong

[Marie Rognes]

Plots of mesh functions over facets look wrong (and sometimes very wrong). A simple example is provided below.

from dolfin import *

mesh = UnitSquare(8, 8)

class Right(SubDomain):
    def inside(self, x, on_boundary):
        return x[0] < DOLFIN_EPS

right = Right()
mf = MeshFunction("uint", mesh, 1)
mf.set_all(0)
right.mark(mf, 1)

plot(mf, interactive=True)

[Anders Logg (logg)]

In the meantime, write to a VTK file and open in Paraview. That should work better.

[Garth Wells (garth-wells)]

This works 'properly' via a VTK file, whereas Viper performs an interpolation which makes the output impossible to properly interpret at boundaries where the MeshFunction value changes. Since it is supported via VTK, the easiest 'fix' would be to disable plotting of MeshFunctions via Viper.

[Anders Logg (logg)]

On Tue, May 03, 2011 at 11:31:22AM -0000, Garth Wells wrote:
> This works 'properly' via a VTK file, whereas Viper performs an
> interpolation which makes the output impossible to properly interpret at
> boundaries where the MeshFunction value changes. Since it is supported
> via VTK, the easiest 'fix' would be to disable plotting of MeshFunctions
> via Viper.

I don't think that's a good fix. I would prefer if plotting
MeshFunctions on facets would issue a warning.

--
Anders

[Joachim Haga (jobh)]

Just a heads-up. This was supposed to be fixed in dolfin with the recent plot-c++ merge, but it still looks wonky. I'll get it sorted out asap.

[Joachim Haga (jobh)]

Fix is pushed to lp:~jobh/dolfin/plot-fix.

[Benjamin Kehlet (benjamik)]

Great! Would you mind (next time) to pull from trunk before fixing. The branch has diverged so manual merging was necessary.