dolfin team mailing list archive

[Anders Logg <question99875@xxxxxxxxxxxxxxxxxxxxx>]

Question #99875 on DOLFIN changed:
https://answers.launchpad.net/dolfin/+question/99875

    Status: Open => Answered

Anders Logg proposed the following answer:
On Thu, Feb 04, 2010 at 12:08:58PM -0000, dbeacham wrote:
> New question #99875 on DOLFIN:
> https://answers.launchpad.net/dolfin/+question/99875
>
> I think I'm right in thinking that dolfin can only perform n-dimensional calculations on meshes embedded in R^n, however is there an easy way to restrict the dof/manipulate the assembly such that when I do something like
>
> a = u*v*ds
> L = f*v*ds
>
> the assembled forms have non-zero determinant? Currently the assembly gives these matrices (reordered):
>       [ds stuff |    0   ]
> A=  [------------------]
>       [      0    |    0   ]
>       [ds stuff ]
> b=  [----------]
>       [      0    ]
>
> whereas I'd either like to just keep the ds stuff (although that would require another mapping between the reordered ds nodes and the original mesh) or possibly set the bottom right of A to the identity matrix (without having to manipulate the matrix myself).
>
> Alternatively, are there any better approaches I could adopt?

I don't know of a way other than to assemble the matrices and then
extract the relevant portions into a new matrix.

We have had some support earlier for restriction of the global set of
dofs to a subset but that has been broken for a while.

--
Anders

-- 
You received this question notification because you are a member of
DOLFIN Team, which is an answer contact for DOLFIN.