[Dune] Re: Notification from Dune

[Oliver Sander]

Hi Robert!
I'm afraid I didn't express myself precisely enough in the
FlySpray entry.  The problem is not the ordering of the
vertices of the reference elements as such but the ordering
of the vertices of subfaces of the reference elements when
you access them through the subEntity method.  Of course
I don't know your code but I'd be surprised if it relied
a lot on that ordering.

Secondly it's not that all normals of the reference quad
point outwards while they should point inwards.  The
problem is that _some_ point outwards and _some_ inwards.
Of course I know which do which, and I can write code
like
if (subface==1 || subface==3)
   // reorient face

That works (and therefore fixing the bug is not urgent for
me).  But it is not nice, and more importantly, it took
quite a while before I found out why my code worked only
on some grids and not on others.

--
Oliver

************************************************************************
* Oliver Sander                ** email: sander at mi.fu-berlin.de        *
* Freie Universität Berlin     ** phone: + 49 (30) 838 75217           *
* Institut für Mathematik II   ** URL  : page.mi.fu-berlin.de/~sander  *
* Arnimallee 6                 ** -------------------------------------*
* 14195 Berlin, Germany        ** Member of MATHEON (www.matheon.de)   *
************************************************************************

On Wed, 18 Oct 2006, Robert Kloefkorn wrote:

> Hi Christian, hi others,
>
>> We discussed this issue in Stuttgart and we decided, that this issue
>> must be solved before 1.0 in the one ore other way. Oliver complains,
>> that the faces in the reference quad are oriented with some normal
>> vectors pointing outwards and some pointing inwards.
>
> The problem described in the task in not nice, I agree.
>
>> We agree, that it would be nice to have all normalvectors pointing
>> inwards (like in the reference triangle). The current orientation
>> arises from the fact, that the reference cubes are constructed via
>> tensor product.
>
> Peter decided to have the reference cubes that way to provide
> n-dimensions reference cubes. That is a good reason.
>
> To keep this feature it would be a better way to change the simplex
> reference elements. It really makes no difference whether the normals
> point outward or inward, you just have to know which direction.
>
>> If it is managable, we should do this change before 1.0, because
>> changing the semantics of the reference cube after 1.0 would be a
>> really bad idea; the api would stay the same and no user whould be
>> warned to change his code. Really strange and hard to find errors
>> would be the result.
>
> This change is gonna be hard at any time. It will effect almost all
> parts of the ALUGrid implementation, the complete dune-fem package,
> everything that has to do with quadrature rules, visualization (at least
> grape). This will change be worse than the GeometryType change.
> I'm, so to speak, "completely against" this change.
>
>> If it is to hard to implement we should just close the bug as "won't
>> fix".
>
> This problem arrises from the definition, that the geometry type line is
> always interpreted as a cube which is (with our current implementation
> of the reference elements) obviously not the case. If one can fix this
> problem, the described problem in the task will disappear.
> An easy fix will be, that a line, belonging to a triangle, is a simplex
> too, a line belonging to a quadrilateral is a cube.
>
> Regards,
>
> R
>
> -- 
>
>  Robert Klöfkorn           <robertk at mathematik.uni-freiburg.de>
>
>  Mathematisches Institut              Tel: +49 (0) 761 203 5631
>  Abt. für Angewandte Mathematik       Fax: +49 (0) 761 203 5632
>  Universität Freiburg
>  Hermann-Herder-Str. 10
>  79104 Freiburg
>
>  http://www.mathematik.uni-freiburg.de/IAM/homepages/robertk
>
> _______________________________________________
> Dune mailing list
> Dune at dune-project.org
> http://www.dune-project.org/cgi-bin/mailman/listinfo/dune
>