[Dune] Comment to task #120.

[Oliver Sander]

Hi!
This sounds like we basically agree that the reference elements
implementation is not optimal.  And within the last 24h i've heard
several proposal for outside-of-the-ReferenceElements-class hacks
that would correct the behaviour.  But even if we invent the
most beautiful hack and put it in grid/utility, it remains a
hack, and a constant reminder that we actually would like the
ReferenceElements class to do something different.  Therefore
I am not in favour of Andreas' proposal and I'd rather simply
document the insufficiencies of ReferenceElements.

--
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, Andreas Dedner wrote:

> Hi,
> in 3d the problem is even greater and i'm sure
> will find no reasonable way to fix the problem in n-d. -
> and not for 1.0 is my opinion.
> For alugrid we fixed the whole think with the following - using
> static variables (vgl. grid/alugrid/3d/topology.cc line 80ff
>
>  template <>
>  const int ElementTopologyMapping<hexa>::
>  dune2aluFaceVertex_[numFaces][numVerticesPerFace] = {{0, 3, 1, 2},
>                                                       {0, 1, 3, 2},
>                                                       {0, 1, 3, 2},
>                                                       {1, 0, 2, 3},
>                                                       {0, 3, 1, 2},
>                                                       {0, 1, 3, 2}};
>
> In 2d it would look simple and one would use:
>  dune2OrientedFace[i][j]
> to get exactly the numbering you need for the correct normal vectors...
>
> This could be easily added to grid/utility for 2d and 3d (and 4d and 5d
> and so one if somebody needs it)
>
> Andreas
>
> PS: the twistutility.hh would not help in this case...
>
>
>
>
> Oliver Sander wrote:
>> Hi Robert!
>>
>>>
>>>> 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
>>>
>>>
>>> if that is the problem then I suggest to add a method to the reference
>>> elementes to return 1.0 or -1.0 depending on the normal pointing outward
>>> or inward. By multiplying with this factor, one always get a outward (or
>>> maybe inward) pointing normal for each face. The we have to change
>>> nothing of our exsiting code and we solve this problem too. Also this
>>> factor tell you the orientation of this face and you can take care of it
>>> in your code.
>>
>>
>> That would not really help much.  If I needed the actual outward
>> normals I would simply call the corresponding method of the
>> IntersectionIterator.  However, I need the grid boundary as a
>> consistently oriented subgrid, so an external library can do
>> whatever with it.  So I'd still need an 'if'-clause like the
>> one above, and since I already know, which faces are incorrectly
>> oriented (faces 0 and 3), supplying that same information
>> in form of a double doesn't really help.
>>
>>
>>> Besides orientation of faces is, what a TwistUtility is good for.
>>> Andreas will comment that piece of code.
>>>
>>
>> Great!  I'm very curious!
>>
>> Oliver
>>
>> _______________________________________________
>> Dune mailing list
>> Dune at dune-project.org
>> http://www.dune-project.org/cgi-bin/mailman/listinfo/dune
>
> _______________________________________________
> Dune mailing list
> Dune at dune-project.org
> http://www.dune-project.org/cgi-bin/mailman/listinfo/dune
>