Detect non 2-manifold meshes in HalfEdgeTopology constructor

[halleysfifthinc]

Background context: I have a non 2-manifold mesh that is happily converted to a HalfEdgeTopology, but calls to loop result in runaway memory usage. It took me a while to discover the root cause (but also spurred the side-quest #1183) since the runaway memory use in loop caused julia to be out-of-memory killed. 🙁

I don't have any alternate design ideas for the loop function, but the unbounded nature is rather unfortunate, since it allows unbounded memory use/growth for poorly behaved meshes.

I don't understand HalfEdge topology conceptually well enough at the moment to know if/how this could be done, but it seems to me that the best solution would be construction time validation of the mesh being 2-manifold.

Of course, this is tangentially related to #354 Repair(2), Repair(4), and Repair(5) which would be great to have but are currently too tangential for me to work on.

[juliohm]

Background context: I have a non 2-manifold mesh that is happily converted to a HalfEdgeTopology

It is an accident that it works. The HalfEdgeTopology is intended for orientable 2-manifold meshes:

Meshes.jl/src/topologies/halfedge.jl

Lines 43 to 49 in 28c7555

	"""
	HalfEdgeTopology(elements; sort=true)
	HalfEdgeTopology(halfedges)
	A data structure for orientable 2-manifolds based on
	half-edges constructed from a vector of connectivity
	`elements` or from a vector of pairs of `halfedges`.

I understand your proposal is to add some sort of check in the HalfEdgeTopology constructor to throw an error when the property is not satisfied.

[halleysfifthinc]

Correct, I had seen the docstring and gathered that my badly behaved mesh was not intended to be constructable/convertible to a HalfEdgeTopology. As such, an error at construction time would be a much better user experience compared to mine, where a later operation ultimately resulted in the Julia process getting killed by the system (which is rather inconvenient for debugging).