Struct for composition?

[antoine-levitt]

We pass around composition a lot and end up documenting it in several places. Should we just give it a struct?

[mfherbst]

That's not a bad idea. I thought about it before and wanted to leave it until we see more clearly how we need it.

One advantage would be that it would be easier to hide some pymatgen details for things such as doubling the unit cell for supercell sampling, which is definitely something we'll be using a lot more in the future.

[antoine-levitt]

• generalize Species to AbstractSpecies, with subtypes (coulomb, psp, ...)
• have a species (Vector of pairs (spec, pos)) field in Model
• make the functions that currently take composition... read from that field
• move the computation of number of electrons to Model (overwritable with a kwarg)

[antoine-levitt]

Why Species and not Atom, btw?

[antoine-levitt]

The reason I'm asking is because with Atom I can figure out what its plural is :D

[mfherbst]

Because in coarse-graining methods the entities of the unit cell might not be atoms, but more generic species :P

[mfherbst]

Plural is boring: species as well ... see wiktionary

[antoine-levitt]

Yeah but species::Vector{Species} is weird.

[mfherbst]

Agreed ... I don't have a better idea at the moment however. I'm also not in favour if "atoms" to be fair.