Re: project progress

[c.]

On 4 Sep 2013, at 14:58, marco Vassallo <address@hidden> wrote:

> Hi,
> I'm sorry to have not yet published anything since your last request but I 
> had an examination this morning and I forgot to warn you, sorry.

No worries, hope you passed the exam ;)

> > - fem-fenics package release
> > - side by side comparison of Octave and Python examples
> > 
> 
> I'm working on these task as you suggested but I've not yet published the 
> code side by side because I 
> still have some doubt about naming convention.
> In particular, at the moment, I create the function for the 
> FunctionalspaceSpace and the Form using
> 
>     fem_create_fs ('problem')
> 
> and it generate the function
> 
>     V = problem_FunctionalSpace(mesh)
> 
> which would like to mimic the c++ command problem::FunctionalSpace.
> I was wondering if it is maybe better to create a standard function 
> FunctionalSpace which receive as input also the name of the problem:
> 
>      fem_create_fs ('problem')
>      V = FunctionSpace (mesh, 'problem')
> 
> Which one do you think it is better? 

I personally like the version with two inputs but I think the really best 
option is the one that would sound more familiar to current FEnics users.
I suggest the following, there are a few FEnics users at MOX which could serve 
as beta testers, if you could come tomorrow or Friday we could discuss this
style issues with them.

> Also, I still don't have figure out how the function which register all the 
> types should work, and so I
> think that I will remove it from the first release of the pkg. (fem_init_env 
> ())
> 
> 
> > Also, do you have any update on the MTW element implementation or the 
> > GenericMatix::nnz method?
> >
> No update on the MTW (and I don't think that anyone will ever do it) and also 
> no update on the nnz
> https://bitbucket.org/fenics-project/dolfin/issue/110/number-of-non-zero-entries-of-a-matrix

that is sad that would have been a great application example, it is beyond the 
scope of your GSoC to extend Dolfin with new elements though so I wouldn't 
worry too much.

On the other hand I am not sure the way your weak form of the Brinkman problem 
is implemented is correct, I would expect a linear system of the form

[A B' 0 ;
 B 0  E';
 0 E  0];

where E is a row vector and derives from the 0-mean constraint ...

> but I will give a better look to this problem later because at the moment, as 
> you suggested, the priority is to put out the fem-fenics pkg to start getting 
> feedback. I think that by the end of the week I should be able to publish all 
> the examples side by side and a first version of the pkg.

yes, I tried to install fem-fenics as a package on an Ubuntu system today and 
the installation failed ...
I'll provide more details later.

> Thanks a lot
> 
> Marco 

Nice work till now, congratulations!
This last phase is crucial though as a lot of clean-up of code and 
documentation is needed to make the package really usable and maintainable.

c.