[Libmesh-users] GMV_IO

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

Hi all

About a week ago I promised to send a patch for the correct export of
QUAD8 elements to GMV. Now I was looking into doing that, and noted
some problems / inconsistencies:

o is mesh_gmv_support.C obsolete? There, the elements are handled
  differently than in gmv_IO (no calls to tecplot_connectivity)

o should there be a gmv_connectivity function, which delegates to
  tecplot_connectivity if there is a special case?
  Or would a generic get_connectivity(se, type) make more sense?
  (where type is one of ["gmv", "tecplot", "diva", "vtk"])
  
o There seems to be quite some code duplication, especially in the
  tecplot_connectivity and vtk_connectivity methods of the different
  geom/cell_*. It seems that the only difference is that gmv needs a
  1 based node numbering.

o Is there a reason for the different calling sequences of the
  tecplot_connectivity and vtk_connectivity methods methods (return
  values vs. return arguments)?

To me it seems that the renumbering / splitting could be handled more
generically. What are your thoughts?

Martin