What type parameters should AbstractSystem have?

[rkurchin]

Currently we have one D for number of dimensions and one S for the species type. There could be more (e.g. adding Unitful types to specify units for length, velocity, etc.) or fewer, i.e. removing one/both of the current ones. Discuss!

[cortner]

Could somebody quickly recap for me why S is required as a type parameter for an abstract system?

[mfherbst]

It turned out that keeping the species type parameter made the implementation of an AtomView very hard, so #34 will drop this type parameter, leaving only the dimensionality as type parameter.

[cortner]

I need a type parameter for the cell, so I just added another abstract type that has {D, TCELL}. No harm.

The bigger question here is why even D? What does D even mean? If my particle has (position, momentum, mass) is D = 3 or is D = 7? Suppose my particles are electrons and I do VMC so I treat them as a particle system and not a field. is the dimension 3 or 4? (spin) I can keep going ...

Could it become problematic that the D for the particles and the D for the cells need to be the same? It seems to be a bit of a question how broad our audience is. If we restrict ourselves to MD then it seems fine.

[jgreener64]

It's the number of dimensions of the space the atoms are in. Generally 3 but could be 2 for a simulation of a surface. So usually the D in the SVector{D, T} for each position, though there is freedom in how to store the positions.

One example use is dispatching for visualisation, where 3D and 2D visualisations would use different code paths.

[cortner]

of the space the atoms are in

... I understand the intention fine. I was just trying to point out that the interpretation is not quite as clear-cut as one might think the moment we go outside the most basic MD type problems which is exactly what I said above.

Again - I think the way we interpret D right now is fine as long as we keep the audience of the interface limited to those that require point particles that have a position variable living in R^D.

[rkurchin]

I agree with Christoph that strictly there is ambiguity (e.g. for spin as you say) but also agree with Joe that for purposes such as dispatching for visualization, it makes sense to have this. Also, for the position/momenta point, this is still consistent/useful because positions/momenta should be of the same dimensions. Similarly, the length of the PBC vector we've been discussing in #97 should be this too...

[cortner]

I think that means we keep what we have now, document clearly what D means and then close the issue?

[rkurchin]

Sounds fine to me – in terms of "document clearly," the docs already say

The D parameter indicates the number of spatial dimensions in the system.

Do you think something beyond this is needed?

[cortner]

To my ears this is ambiguous for the reasons I explained above. "Space " for a mathematician and actually many physicists as well is a much broader concept.

[rkurchin]

How about adding a parenthetical along the lines of "(i.e., the dimension of the position vectors)"?

[mfherbst]

I'll close this due to the merged PR. Feel free to reopen if needed.