ex13: add mesh.refined(1) and everything is broken

[gatling-nrl]

Add mesh = mesh.refined(1) before line 40

scikit-fem/docs/examples/ex13.py

Lines 36 to 40 in 71eb03e

	mesh = mesh.translated(mesh.p[0] * np.stack([np.cos(mesh.p[1]),
	np.sin(mesh.p[1])]) - mesh.p)
	elements = ElementTriP2()
	basis = Basis(mesh, elements)

and you get

---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
~\AppData\Local\Temp/ipykernel_21440/3992918866.py in <module>
     32 
     33 u = basis.zeros()
---> 34 u[basis.get_dofs("positive")] = 1.
     35 u = solve(*condense(A, x=u, D=basis.get_dofs({"positive", "ground"})))
     36 

c:\users\username\local\scikit-fem\skfem\assembly\basis\abstract_basis.py in get_dofs(self, facets, elements, skip)
    243                                               skip_dofnames=skip)
    244 
--> 245         facets = self._get_dofs_normalize_facets(facets)
    246         return self.dofs.get_facet_dofs(facets,
    247                                         skip_dofnames=skip)

c:\users\username\local\scikit-fem\skfem\assembly\basis\abstract_basis.py in _get_dofs_normalize_facets(self, facets)
    261             return self.mesh.facets_satisfying(facets)
    262         elif isinstance(facets, str):
--> 263             return self.mesh.boundaries[facets]
    264         raise NotImplementedError
    265 

TypeError: 'NoneType' object is not subscriptable

[gdmcbain]

Ouch! #104?

[kinnala]

This is already fixed in #792, just waiting for it to be merged.

Edit: However, it is not fixed for all Mesh types.

We used to have a warning if boundaries and subdomains are invalidated by a refine but I have forgotten to include it in the large Mesh refactor in 3.0.0. I'll try adding it again.

Also some kind of proper error message would be nice if the boundary name does not exist when calling get_dofs.

[kinnala]

Let me add that currently subdomains are always invalidated by a refine.

[kinnala]

I did some changes in #792 and now it looks like this if refined does not preserve the names:

In [1]: from skfem import *                                                   

In [2]: m = MeshTet().refined()                                               
Named boundaries invalidated by a call to Mesh.refined()

In [3]: basis = Basis(m, ElementTetP1())                                      

In [4]: basis.get_dofs("left")                                                
---------------------------------------------------------------------------
ValueError                                Traceback (most recent call last)
<ipython-input-4-9d9e979f36a3> in <module>
----> 1 basis.get_dofs("left")

~/src/scikit-fem/skfem/assembly/basis/abstract_basis.py in get_dofs(self, facets, elements, skip)
    243                                               skip_dofnames=skip)
    244 
--> 245         facets = self._get_dofs_normalize_facets(facets)
    246         return self.dofs.get_facet_dofs(facets,
    247                                         skip_dofnames=skip)

~/src/scikit-fem/skfem/assembly/basis/abstract_basis.py in _get_dofs_normalize_facets(self, facets)
    265                 return self.mesh.boundaries[facets]
    266             else:
--> 267                 raise ValueError("Boundary '{}' not found.".format(facets))
    268         raise NotImplementedError
    269 

ValueError: Boundary 'left' not found.

The ultimate goal, of course, is to have these preserved for all Mesh types.