def test_merge_points_large(dim, radius):
points = 10 * torch.randn(200, dim)
merged_points, groups = merge_points(points, radius)
logger.debug('merged {} -> {}'.format(len(points), len(merged_points)))
assert merged_points.dim() == 2
assert merged_points.shape[-1] == dim
assert len(groups) == len(merged_points)
assert sum(len(g) for g in groups) == len(points)
assert set(sum(groups, ())) == set(range(len(points)))
d2 = (merged_points.unsqueeze(-2) - merged_points.unsqueeze(-3)).pow(2).sum(-1)
assert d2.min() < radius ** 2
def test_merge_points_small():
points = torch.tensor([
[0., 0.],
[0., 1.],
[2., 0.],
[2., 0.5],
[2., 1.0],
])
merged_points, groups = merge_points(points, radius=1.0)
assert len(merged_points) == 3
assert set(map(frozenset, groups)) == set(map(frozenset, [[0], [1], [2, 3, 4]]))
assert_equal(merged_points[0], points[0])
assert_equal(merged_points[1], points[1])
assert merged_points[2, 0] == 2
assert 0.325 <= merged_points[2, 1] <= 0.625