My first thought is that lattice is nicer than supercell, it feels more natural 👍

Then there could also be a Sites class and a Geometry would be the combination of Lattice, Sites and Atoms. A combination of only Sites and Lattice could implement the purely geometrical methods that are now implemented in Geometry, which maybe would be useful for other things. The first thing that comes to mind is k points in the reciprocal lattice, but I'm sure it could have other applications e.g. in situations where you have a sparse sampling of space :)

Codecov Report

Merging #550 (2d45aad) into main (1a08992) will increase coverage by 0.03%.
The diff coverage is 94.30%.

@@            Coverage Diff             @@
##             main     #550      +/-   ##
==========================================
+ Coverage   86.20%   86.23%   +0.03%     
==========================================
  Files         370      370              
  Lines       46702    46777      +75     
==========================================
+ Hits        40258    40338      +80     
+ Misses       6444     6439       -5     

Help us with your feedback. Take ten seconds to tell us how you rate us. Have a feature suggestion? Share it here.

I like Lattice too, as well as @pfebrer's suggestion of Sites.

Note that in 3D there exists only 14 different Bravais lattices. Maybe they could be introduced for some symmetry-related functionality?

Ok, so let's merge this and be gone with the madness. @tfrederiksen once merged, it would be nice to see if hubbard works with no changes, I have tried to catch tests as much as possible, but with this major change I have the feeling that I forgot something ;)

With inspiration from spglib, one could rename .cell to .basis (basis vectors) and reserve cell for information about the connectivity etc.

Hmm, I am not too fond of that convention. Kind of like cell like unit-cell. I don't fully get your last paragraph?

I was considering abc or cell, I think basis is a little more "mathy", and not so physics oriented. Could be wrong though?

I am just wondering if the notion of nsc is clear for the user, and if basis would be a more natural property of Lattice.

But I'm also fine with cell as it is.

agreed, nsc could probably be spelled out supercells. I just hate long words... ;)

Isn't it rather that a supercell means the collection of the unit cell + its neighbours?

yes, but that is essentially what nsc means, no? Or should we do nsc = cell_neighbours, or something similar? The supercell phrase is just used extensively (in the code and in literature).

Another word could be connections.

That's not entirely fitting either, is it orbital connections, or ...

Hmm... in the sparse structures we use Sparse.edges but neither are referring to the clarity of supercells, I think. The fact that this is an auxiliary supercell I think is quite vital in the understanding of its use, and if that gets lots in the name, I am afraid it will confuse?

It is weird because nsc is not really a property of the lattice, no? It only makes sense in a combination of a lattice, some sites, and something that defines the range of the sites. Perhaps the lattice information should be separated from the auxiliary cell information, which could be an attribute -aux_cell- of Geometry or whatever other class where it makes sense.

So lattice would hold cell and pbc, and aux_cell would hold nsc.

Supercells in plural is what I find a bit confusing terminology.

It is weird because nsc is not really a property of the lattice, no? It only makes sense in a combination of a lattice, some sites, and something that defines the range of the sites. Perhaps the lattice information should be separated from the auxiliary cell information, which could be an attribute -aux_cell- of Geometry or whatever other class where it makes sense.

So lattice would hold cell and pbc, and aux_cell would hold nsc.

True... The plot thickens...