Anders Logg wrote:
> On Tue, Jun 23, 2009 at 09:25:19AM +0200, Kent Andre wrote:
>   
>> On ma., 2009-06-22 at 20:57 +0200, Anders Logg wrote:
>>     
>>> On Mon, Jun 22, 2009 at 12:46:09PM -0600, Bartosz Sawicki wrote:
>>>       
>>>> Kent Andre wrote:
>>>>         
>>>>> On ma., 2009-06-22 at 00:40 -0600, Bartosz Sawicki wrote:
>>>>>           
>>>>>> On 22/06/09 12:06 AM, Anders Logg wrote:
>>>>>>             
>>>>>>> On Sun, Jun 21, 2009 at 11:36:29PM -0600, Bartosz Sawicki wrote:
>>>>>>>               
>>>>>>>> On 19/06/09 02:53 PM, kent-...@simula.no wrote:
>>>>>>>>                 
>>>>>>>>> When using VMTK the boundary indicators are stored as a set of arrays
>>>>>>>>> with names
>>>>>>>>> boundary facet cells,
>>>>>>>>> boundary facet numbers
>>>>>>>>> boundary indicators
>>>>>>>>> The two first arrays are used to find the global facet numbers.
>>>>>>>>> The function computes a mesh function based on these arrays.
>>>>>>>>>
>>>>>>>>> This was what Anders prefered since VMTK does not now anything
>>>>>>>>> about the global numbering of facets in Dolfin (without linking it 
>>>>>>>>> into
>>>>>>>>> Dolfin).
>>>>>>>>>
>>>>>>>>> This is not a VMTK specific problem, but I do think VMTK is the only
>>>>>>>>> mesh-generator
>>>>>>>>> that let you store boundary indicators at the moment. I guess we 
>>>>>>>>> should add
>>>>>>>>> this to NetGen as well, both Hans Petter Langtangen and Douglas 
>>>>>>>>> Arnold has
>>>>>>>>> bugged
>>>>>>>>> me about this the last few days.
>>>>>>>>>                   
>>>>>>>> I'm every day user of this functionality. I use Netgen generated 
>>>>>>>> models, 
>>>>>>>> but also other software which add boundary indicators to existing 
>>>>>>>> meshes. Netgen supports boundary indicators quite well, both by geo 
>>>>>>>> files, and by GUI. I don't understand what you want to add?
>>>>>>>>                 
>>>>>>> Is it possible to set DOLFIN boundary indicators from Netgen? I didn't
>>>>>>> know.
>>>>>>>               
>>>>>> It is definitely possible :) I don't have direct converter from Netgen 
>>>>>> vol to Dolfin xml. Most of my old models were stored in Diffpack grid 
>>>>>> format, so I'm using grid to xml converter.
>>>>>> Netgen can export to the grid format, so this is not a big issue.
>>>>>>             
>>>>> It is not possible to store the boundary indicators in Dolfin format
>>>>> using Netgen, at  the current time. I wrote the Netgen -> dolfin filter
>>>>> in Netgen so I should know :).  
>>>>>
>>>>> But it should be possible to include support for that in a similar
>>>>> fashion as we did in VMTK. 
>>>>>           
>>>> You are right that Netgen doesn't support Dolfin XML mesh format. But it 
>>>> doesn't mean that is not possible to use boundary indicators generated 
>>>> by Netgen in Dolfin. Converterting via Diffpack grid file preserves 
>>>> boundaries.
>>>>
>>>> The algoritm is simple:
>>>> 1. Generate mesh using Netgen.
>>>> 2. Mark boundaries indicators.
>>>> 3. Export to Diffpack format.
>>>> 4. Convert Diffpack to XML.
>>>> 5. Enjoy boundary indicators in Dolfin.
>>>>
>>>> cheers,
>>>>         
>>> Nice, but not optimal. At some point, we need an application that can
>>> do all of these (and more). There seem to be a couple of different
>>> options:
>>>
>>>   1. Extend Netgen with full support for DOLFIN (and change the toolkit...) 
>>>   2. Extend Gmsh with full support for DOLFIN (and change the toolkit...)
>>>   3. Extend MeshBuilder to do mesh generation (using for example Netgen)
>>>   4. ?
>>>
>>> Any thoughts on these options?
>>>
>>> The problem with (3) is that no one is really developing MeshBuilder,
>>> except when we pay Kalkulo (http://simula.no/applications/kalkulo/) to
>>> do it.
>>>
>>>       
>> It should be fairly easy to include bc indicators in NetGen. I can look
>> at it, I think I need to update the filter anyway to make sure the mesh
>> is ordered correctly according to UFC. 
>>
>> In general it is rather easy to write out the mesh and indicators
>> compared to actually making it. I bet we can make Gmsh write Dolfin mesh
>> in just a few hours (but I have not looked at Gmsh). 
>>
>> The Dolfin XML file format needs to be documented. 
>>     
>
> Yes, and there's some ongoing work on creating schemas and validating
> against these schemas when parsing (Relax-NG using libxml2).
>
> Ilmar has been working on it but I don't know the current
> status. Ilmar? Ola?
>
>   
The functions for validating the schemas are in place, the use the 
function validate in NewXMLFile.cpp.

The schema file is available on the fenics server as pub/misc/dolfin.rng 
now, and is used in the validate function. We might want to add a 
release number to this file, e.g. dolfin-0.9.2.rng, as the schema might 
change as new mesh types are added.

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

Reply via email to