Re: [Libmesh-users] Data projection on mesh refinement

[<an...@tn...>]

Hi Steffen

Thanks for your answer. Unfortunately I think it doesn't answer my question.

Martin writes:
 > >Upon mesh refinement, the mesh.data is not projected on the new mesh.
 > >To my understanding this could be accomplished in a similar way as the
 > >projection of solution vectors. Indeed the only workaround I could
 > >come up with was to add a trivial system, solve it for the nodal data
 > >on the RHS, and subsequently use the solution of this system to
 > >obtain the data values on a refined mesh.

Steffen Petersen writes:
 > Try the MeshData::translate function. Than you do not have to solve a system
 > to get the data in a vector, and let the system do the projection. I am 
 > not quite sure
 > if a function similar to project_vector is really necessary in MeshData 
 > (code duplication!?).

Well, looking at the code I don't think that the code does
that. MeshData.translate seems to be useful to extract data on the
current mesh. What I want is:

o read data from a file into a MeshData object
o access the same data on a refined mesh (e.g. in an element loop)
  with the data interpolated to the new nodes (of the refined mesh)

As proposed in the original message, I still think that the MeshData
should have a reinit method that calls a project_vector on every data
vector.

BTW: the name "MeshData::translate" is quite confusing, especially
since the Mesh class has a translate method which actually translates
the mesh in space. Would "MeshData::get_data_vectors" be a more
appropriate name?


Best, Martin