Improve documentation to calculate metrics from just generated mesh
Closed this issue · 4 comments
The notebook for mesh metrics show it with a mesh that it loads from file. Now I try to compute them from a just generated mesh without saving and loading it. The basic approach from the script is
# Image processing happening before this
mesher = Mesher2D(segmentation)
mesher.generate_contour(max_contour_dist=3)
# Some padding and setting region markers left out
mesh = mesher.triangulate(opts='pAq30a')
metrics.histogram(mesh, metric='max_angle')
# ax = metrics.plot2d(mesh, metric='max_angle')
As result it errors on making the histogram
Warning: Appending zeros to replace the missing geometrical tag data.
---------------------------------------------------------------------------
AttributeError Traceback (most recent call last)
/var/folders/yp/w25rz7hs47n4mhxgqmbl8xch0000gn/T/ipykernel_64069/1301693132.py in <module>
6 fig.set_size_inches(24,4)
7
----> 8 metrics.histogram(mesh_local_thresh_wave, metric='max_angle')
9 ax = metrics.plot2d(mesh_local_thresh_wave, metric='max_angle')
/usr/local/Caskroom/miniconda/base/envs/nanomesh/lib/python3.8/site-packages/nanomesh/metrics.py in histogram(mesh, metric, ax, **kwargs)
296
297 descriptor = _metric_dispatch[metric]
--> 298 quality = descriptor.func(mesh) # type: ignore
299
300 fig, ax = plt.subplots()
/usr/local/Caskroom/miniconda/base/envs/nanomesh/lib/python3.8/site-packages/nanomesh/metrics.py in __call__(self, mesh)
23
24 def __call__(self, mesh: BaseMesh) -> np.ndarray:
---> 25 grid = mesh.to_pyvista_unstructured_grid()
26 ret = grid.compute_cell_quality(self.metric)
27 quality = ret.cell_data['CellQuality']
AttributeError: 'MeshContainer' object has no attribute 'to_pyvista_unstructured_grid'
This is working as intendend. You must specify which cell type you want to run the metrics on.
The triangulation always returns a MeshContainer, which includes line and triangle cells.
Try this:
triangle_mesh = mesh.get('triangle')
metrics.histogram(triangle_mesh, metric='max_angle')That indeed works. Could it be useful to highlight/document this difference?
As idea, the metrics example starts with:
mesh = TriangleMesh.from_meshio(meshio.read('out.msh'))
mesh.prune_z_0()
mesh.plot()
....
metrics.histogram(mesh, metric='min_angle')
Which might not make it directly clear that TriangleMesh is quite different from the MeshContainer. The latter does contain multiple meshes, but in practice I do think of it like the mesh for the highest dimension.
Yeah, good point. That notebook needs some attention anyway. I don't think I updated it since I first made it 😅