Hello,

I get the error " ImportError: cannot import name _from_cell_list " when running the following code (the mesh file can be found attached):

from firedrake import *
import numpy as np

mesh_data = np.loadtxt("sweVbuoy2D_mesh.in")

xcoords = mesh_data[:,[0,2,4,6]]
ycoords = mesh_data[:,[1,3,5,7]]
coords = np.asarray(np.meshgrid(xcoords, ycoords)).swapaxes(0, 2).reshape(-1, 2)

cells = mesh_data[:,[8,9,10,11]].astype(int)
#cells = np.asarray(cells).swapaxes(0, 2).reshape(-1, 4)

from firedrake.mesh import _from_cell_list
dm = _from_cell_list(2, cells, coords)
mesh = Mesh(dm)

V = FunctionSpace(mesh, "CG", 1)
f = Function(V)

f_file = File("f.pvd")
f_file << f


Anyone has any idea why this is happening?

Thanks, Anna.


On 16/12/15 19:43, Lawrence Mitchell wrote:

On 16 Dec 2015, at 19:07, Anna Kalogirou <a.kalogirou@leeds.ac.uk> wrote:

Thank you very much guys,

So the mesh I already have was constructed manually and has the following format (please also see attached):
it has as many number of lines as the number of elements, with 12 values in each line.
The first 8 values consist of 4 sets of coordinates (x,y) (without the brackets and comma, just 8 numbers)
and the final 4 denote the global node number of the 4 points.

Is it possible to build a mesh on Firedrake based on this file?

If you do:

from firedrake.mesh import _from_cell_list

Then you should be able to use that. It expects you to tell it the topological dimension of your cells, pass a 2d numpy array with the coordinates of your vertices. And a 2d numpy array indicating which vertices belong to which element. You can pass the object you get back to Mesh:

dm= _from_cell_list(2, cells, vertices)

mesh = Mesh(dm)

There are some further examples using this in utility_meshes.py

The format sounds sufficiently simple that you could add a reader in firedrake and we could incorporate it.

Thanks,

Lawrence


On 16/12/15 16:58, Justin Chang wrote:
Anna,

The first four entries with "15" under the column refer to the geometric vertices.

The next 28 with "1' under the second column refer to the mesh straight lines aligning under the geometric straight lines connecting the aforementioned points

The quadruplet information is described per the link Lawrence pointed to.

----

So I see that your mesh has no physical IDs (the "0" under the fourth column), which is needed for imposing boundary conditions. If this were done entirely with GMSH, they order the physical ID's based on the order you create them. If you don't do any of this, GMSH spits out every single geometric and mesh entity with a 0 ID. That's why you see some useless element entries like the four "15" vertices.

Thanks,
Justin

On Wed, Dec 16, 2015 at 7:45 AM, Lawrence Mitchell <lawrence.mitchell@imperial.ac.uk> wrote:
Dear Anna,

On 16/12/15 15:17, Anna Kalogirou wrote:
> Dear all,
>
> I need to import my own mesh which was not created using Gmsh, so I am
> trying to convert it to Gmsh format. I have some questions and I didn't
> find any documentation on this on the Gmsh website.

Before you try and do this conversion, where does the mesh come from?
 PETSc (and therefore we) claim to support the following formats read
from file:

- gmsh (extension .msh)
- exodus (extension .e or .exo)
- cgns (extension .cgns) Although boundary markers aren't supported
- triangle (extension .node)

We also, although it's not a public API, support building a mesh from
a list of cells and their vertex coordinates.

Would any of these suit you?

> For example, in the sample msh file attached there should be 40 elements
> and not 72. I don't understand what the first 32 lines (i.e. lines 64-95
> in the file) under "Elements" actually are. Also, in the actual elements
> (lines 96-135), what is the first quadruplet {3 2 0 5} after the global
> element number?

The documentation for the gmsh file format is
http://www.geuz.org/gmsh/doc/texinfo/gmsh.html#MSH-ASCII-file-format

Note that not all "elements" in the file are what you think of as
elements, some of them will be lines and points I think.

Lawrence


_______________________________________________
firedrake mailing list
firedrake@imperial.ac.uk
https://mailman.ic.ac.uk/mailman/listinfo/firedrake




_______________________________________________
firedrake mailing list
firedrake@imperial.ac.uk
https://mailman.ic.ac.uk/mailman/listinfo/firedrake


-- 
 
 Dr Anna Kalogirou
 Research Fellow
 School of Mathematics
 University of Leeds

 http://www1.maths.leeds.ac.uk/~matak/
<sweVbuoy2D_mesh.in>
_______________________________________________
firedrake mailing list
firedrake@imperial.ac.uk
https://mailman.ic.ac.uk/mailman/listinfo/firedrake


_______________________________________________
firedrake mailing list
firedrake@imperial.ac.uk
https://mailman.ic.ac.uk/mailman/listinfo/firedrake