Re: blog update #2

[c.]

On 17 Jun 2013, at 23:29, c. <address@hidden> wrote:

> Hi Marco,
> 
> On 17 Jun 2013, at 23:05, Marco Vassallo <address@hidden> wrote
> 
>> Hi, I'm trying to add the data stored in the dolfin xml file to the (p,e,t) 
>> matrix.
>> I have some questions about how the code should be designed:
>> 1)
>> Create p,e,t
>> if( additional information available)
>>  add extra information
>> In this case I have to store an extra map which stores the information
>> number_of_dolfin_face_idx -> number_of_octave_face_idx
> 
> 
> I'd use a minimalistic approach at this point: 
> 
> 1) look for subdomain info in the dolfin mesh and fill the last raw of "t"
> 2) lok for boundary subregion info and fill the "e" structure according to 
> the msh format
> 
> at a later point you could also try and fill the extra topological and 
> geometric property fields
> supported by msh and described in the docs for the functions 
> "msh2m_topological_properties" and
> msh2m_geometrical_properties, but that is not a priority at the moment.
> 
> The map you have in mind, in particular, sounds like the "boundary" field 
> computed by "msh2m_topological_properties", 
> is that the case?
> 
>> 2)
>> If additional information is stored
>>   ( create p,e,t, in some way)
>> else
>>   (create p,e,t, in another way)
>> 
>> Is there any recommended way for optimization/clarity?
> 
> well, if no info about bulk/boundary subdomains is present you could use some 
> default choice, 
> e.g., set the boundary mark to 1 an all boundary facets and the subdomain 
> marker to 1 on all elements.
> 
>> Marco
> 
> HTH,
> c.

Marco, 

BTW I followed this thread on the FENICS mailing list [1].

I disagree with the idea that the only possible approach is that of 
creating a library of pre-built variational forms (thaough such a library, 
for common problems would be very useful in my opinion).
I think the problem here is that everyone in that thread assumes we should be
mimicking the python interface, but I think it would be much simpler to try to 
replicate the C++ interface by creating a script to build .oct shared objects 
files from .ufl
files.

HTH,
c.


[1] http://fenicsproject.org/pipermail/fenics/2013-May/000280.html