Is the reordering check for
quad elements completed in
dolfin - 2019.1.0? The corresponding issue is marked as
resolved. However I seem to get an error trying to import
quad meshes (and even when trying to manually create one using
MeshEditor). It is also reported here
from dolfin import *
msh = Mesh()
f = XDMFFile(MPI.comm_world,'test_quadMesh.xdmf')
Attached is a picture of the sample mesh, if needed.
Manually creating the mesh using
nodes, cells = return_points_cells(abaqus_fname)
num_nodes = nodes.shape
num_elems = cells.shape
msh = Mesh()
e = MeshEditor()
for i, nval in enumerate(nodes):
for elem in cells:
e.add_cell(elem,np.array([elem, elem, elem, elem],int))
with XDMFFile(comm,'trialMesh.xdmf') as wf:
results in the following mesh:
It does not read the complete mesh, however, at least does better than throwing a
Hi ! I ran into this as well ! Have you managed to solve that issue ?
Hi, I think this will be (or is already) fixed in the
There is a lot of work going on in dolfinx in quads/hexes. The idea there is to permutation the dofmap instead of ordering the mesh. I think that the permuting of quad dog maps will appear in a PR in dolfinx later this week and be merged early next week.
So this won’t be fixed in the current FEniCS ?
I can’t guarantee that it will. Most development of the FEniCS source code is happening in dolfinx (which hopefully will have a stable release during the first quarter of next year).
Pull request for dolfinx is now submitted: https://github.com/FEniCS/dolfinx/pull/654
Some minor things has to be fixed before its merged, and alleviates the ordering issue for quads and hexes for scalar function-spaces. The next step is to work on permutations of vector spaces instead of ordering the mesh. I guess @mscroggs can give more details