Christoph Ortner

To my mind this sort of interface necessarily involves neighbourlists. I thought it is the responsibility of the neighbourlist to provide lists of bonds - ideally lazily since anything beyond...

> In molecular mechanics terms the bonds are defined during setup and fixed throughout the whole simulation. Thank you for clarifying this for me. If that's the concrete context, I...

Despite resistance to my comments above - I still think a list of bonds and a neighbourlist are very closely related objects (connectivity). It would be great to give some...

> new example system type Why can't this be stored in the meta-data of the existing systems? If there are no bonds stored in the meta-data then the call to...

PLEASE PLEASE do not re-implement the neighbourlist for the 5th time? Can we rather talk about integrating one or all of the existing ones?

> I agree that bonds and neighbor lists are related conceptually, but I think the way they are used in actual simulation (and also in visualization, which is the particular...

The justification for a separate `BondedSystem` type is probably that you are then signalling that the bonds remain fixed. Is that the intention? This makes sense to me. But to...

> stability of a bond representation then you have only pairs. Sure, that's fine. Except don't use `Symbol`, that's type-stable not `isbits`.

> As for neighbor lists, this is something that is really only used in a molecular dynamics simulation and is not really a property of an atomic system whereas bonds...

I'm not trying to be contrary for fun here and as soon as I see a convincing argument that a list of bonded pairs is fundamentally something different from a...