def test_point_skew_2d(niters=1, **kwargs):
'''Not necesarily conform'''
mesh1d = _1d2d_mesh(4)
embedding = embed_mesh1d(mesh1d,
bounding_shape=0.1,
how='as_points',
gmsh_args=[],
niters=niters,
debug=False,
save_geo='',
**kwargs)
skewed = embedding.nc_edge_encoding.as_vertices
assert skewed
x = embedding.embedding_mesh.coordinates()
embedding.embedding_mesh.init(1, 0)
embedding.embedding_mesh.init(0, 1)
E2V, V2E = embedding.embedding_mesh.topology()(1, 0), embedding.embedding_mesh.topology()(0, 1)
compute_star = lambda v: set(np.hstack([E2V(ei) for ei in V2E(v)]))
y = mesh1d.coordinates()
e2v = mesh1d.topology()(1, 0)
for edge in skewed:
y0, y1 = y[e2v(edge)]
for piece in skewed[edge]:
if not len(piece) == 3:
continue
x0, x2, x1 = x[piece]
# End points were inserted correctly
assert is_on_line(x0, y0, y1)
v0, v2, v1 = piece
# Mid point is in the star
mids = compute_star(v0) & compute_star(v1)
assert v2 in mids
# If there are more this way we get a shortest path
if len(mids) > 1:
path_length = lambda v: np.linalg.norm(x[v0] - x[v], 2) + np.linalg.norm(x[v1] - x[v], 2)
assert v2 == min(mids, key=path_length)
vmap = embedding.vertex_map
encode = embedding.edge_encoding.as_edges
# We can combine the maps to pick correctly embedded edges
for edge in range(mesh1d.num_cells()):
if edge not in skewed:
v0, v1 = e2v(edge)
e_ = encode[edge]
print e_, '<--'
vs = set(vmap[e2v(e_[0])])
assert v0 not in vs
vs = set(vmap[e2v(e_[-1])])
assert v1 in vs
def _embed_edgeencode(line_mesh, embedding_res):
'''Vertices on edge are really there'''
x = embedding_res.embedding_mesh.coordinates()
line_mesh.init(1, 0)
e2v = line_mesh.topology()(1, 0)
for edge_idx, edge in enumerate(embedding_res.edge_encoding.as_vertices):
# In embedding_mesh numbering
v0, v1 = edge[0], edge[-1]
# Grab it from line_mesh
V0, V1 = embedding_res.vertex_map[e2v(edge_idx)]
assert v0 == V0 and v1 == v1
# The points are indeed on the edge
edge_x = x[edge]
A, B = edge_x[0], edge_x[-1]
assert all(is_on_line(P, A, B) for P in edge_x)
return True
def test_point_skew_stl_3d():
'''Not necesarily conform'''
mesh1d = _1d3d_mesh(4)
embedding = embed_mesh1d(mesh1d,
bounding_shape='box.stl',
how='as_points',
gmsh_args=[],
niters=1,
debug=False,
save_geo='')
skewed = embedding.nc_edge_encoding.as_vertices
assert skewed
x = embedding.embedding_mesh.coordinates()
embedding.embedding_mesh.init(1, 0)
embedding.embedding_mesh.init(0, 1)
E2V, V2E = embedding.embedding_mesh.topology()(1, 0), embedding.embedding_mesh.topology()(0, 1)
compute_star = lambda v: set(np.hstack([E2V(ei) for ei in V2E(v)]))
y = mesh1d.coordinates()
e2v = mesh1d.topology()(1, 0)
for edge in skewed:
y0, y1 = y[e2v(edge)]
for piece in skewed[edge]:
assert len(piece) == 3
x0, x2, x1 = x[piece]
# End points were inserted correctly
assert is_on_line(x0, y0, y1)
v0, v2, v1 = piece
# Mid point is in the star
mids = compute_star(v0) & compute_star(v1)
assert v2 in mids
# If there are more this way we get a shortest path
if len(mids) > 1:
path_length = lambda v: np.linalg.norm(x[v0] - x[v], 2) + np.linalg.norm(x[v1] - x[v], 2)
assert v2 == min(mids, key=path_length)