`Interchange.from_openmm` should raise an exception if the topology doesn't match the system about openff-interchange HOT 3 CLOSED

How stringent a check are we talking here? The most naive check that comes to mind is checking that the number of particles in the system matches the number of atoms in the topology. (This could be massaged slightly to only consider massive particles, since virtual sites don't exist in OpenFF's topologies.) I imagine that should cover most cases of accidentally passing in the wrong variables.

That sounds great. openmm.System seems to have a isVirtualSite(index) method, which might be preferable to checking mass because OpenMM's recommended way to freeze atoms is to set their mass to zero.

I agree connectivity has all sorts of bugbears that are likely to get worse in the future!

from openff-interchange.

I imagine virtual sites might make this a pain.

Your imagination is correct - but that's implicitly unsupported right now so it's worthwhile to make it explicitly unsupported.

How stringent a check are we talking here? The most naive check that comes to mind is checking that the number of particles in the system matches the number of atoms in the topology. (This could be massaged slightly to only consider massive particles, since virtual sites don't exist in OpenFF's topologies.) I imagine that should cover most cases of accidentally passing in the wrong variables.

I think connectivity checks would be trickier to implement (rigid water often doesn't have bonds in OpenMM models, for example) and risk some false negatives, so I don't think I'd go for that in the first pass here.

from openff-interchange.

#923 & 0.2.23

from openff-interchange.