Small questions about pyvista.Plotter() and FEniCS-X

Yingqi_Jia · May 5, 2021, 5:09am

Hi all! I would like to know how to change the window size of pyvista.Plotter()?
I try to use plotter = pyvista.Plotter(window_size=[600,2000]), and the window size does become larger, but there is a very thick gray edge. How to eliminate the thick gray edge?

In addition, I have two small questions about FEniCS-X:

(1) How to refine the mesh of the quadrilateral elements? It seems the refine function is only applicable to the simplex elements.

(2) How to conduct the parallel computation with MPI.SUM, MPI.MIN, MPI.MAX and so on? In the classical FEniCS, we can use comm = mesh.mpi_comm() and MPI.min(comm, some_value), but it does not work in FEniCS-X.

Thanks!

dokken · May 5, 2021, 5:51am

If you use the interactive mode with pyvista you can scale the window as you please. If you use the screenshot mode, you need to fiddle with Camera parameters. See for instance https://github.com/FEniCS/dolfinx/blob/main/python/demo/pyvista/demo_pyvista.py#L105 and the following lines, or the pyvista documentation.

There is currently no support for refinement of non-simplices.

See for instance:

github.com

FEniCS/dolfinx/blob/bc894b5ab586617caac779a4eba59193a939b7f1/python/test/unit/geometry/test_bounding_box_tree.py#L183


      
                  assert entities_B == references[i][1]
          
          
          @pytest.mark.parametrize("dim", [0, 1])
          def test_compute_closest_entity_1d(dim):
              ref_distance = 0.75
              p = numpy.array([-ref_distance, 0, 0])
              mesh = UnitIntervalMesh(MPI.COMM_WORLD, 16)
              tree = BoundingBoxTree(mesh, dim)
              entity, distance = compute_closest_entity(tree, p, mesh)
              min_distance = MPI.COMM_WORLD.allreduce(distance, op=MPI.MIN)
              assert min_distance == pytest.approx(ref_distance, 1.0e-12)
          
              # Find which entity is colliding with known closest point on mesh
              p_c = numpy.array([0, 0, 0])
              entities = compute_collisions_point(tree, p_c)
          
              # Refine search by checking for actual collision if the entities are
              # cells
              # NOTE: Could be done for all entities if we generalize
              # select_colliding_cells to select_colliding_entities

or the general mpi4py documentation: Overview — MPI for Python 3.1.5 documentation
as dolfinx uses mpi4py.

Yingqi_Jia · May 5, 2021, 6:59am

Thank you, Dokken. And I have another small question about FEniCSX. Is there some way to calculate the gradient (like strain and heat flux) within the domain instead of the nodal points? Since I want to calculate the average gradient over the domain.

dokken · May 5, 2021, 7:17am

Project the gradient into an appropriate space, and use point evaluation (as shown in the membrane tutorial)
You can also have a look at: FEniCS 2021: FFCx code generation for expressions

Yingqi_Jia · May 5, 2021, 7:36am

Yes, actually I was also using this projection method. However, this tutorial only computes the nodal values with pressure.compute_point_values(). Can we compute the values within the element such as the Gauss integration points or center point? Since gradients might be discontinuous at the nodal points.

dokken · May 5, 2021, 9:07am

As shown here: Implementation — FEniCS-X tutorial
You can evaluate a dolfinx function at any point (here sampling a line through) the domain, that is not aligning with grid nodes

Topic		Replies	Views
Error with pyvista in new fenicsx version I/O	7	743	August 31, 2023
Matplotlib instead of pyvista in Dolfinx General	7	1620	April 13, 2023
Pyvista mesh plot not working with fenicsx General	2	877	October 12, 2022
Matplotlib plotting backend doesn't handle quadrilateral mesh mesh	1	989	March 5, 2022
Pyvista doesn't work in pycharm General visualization	8	447	January 8, 2024

Small questions about pyvista.Plotter() and FEniCS-X

Related topics