Default to lattice from trajectory for analysis

[stefsmeets]

SitesData should prefer the lattice from the trajectory instead of model structure. i.e. just load fractional coords from ideal structure.

We could also make it so that the lattices must be the same, so that the user needs to take care of this. Alternative, perhaps scale input ideal lattice to trajectory to enforce lattices are the same.

[stefsmeets]

Hi @AstyLavrinenko , I'm looking at this, but I don't see any instances where it takes the structure lattice over the trajectory lattice. Do you have an example where this goes wrong?

[AstyLavrinenko]

Hi @stefsmeets,

The issue isn't specifically tied to SitesData. I went through sites.py and transitions.py, and it seems like the trajectory lattice is being employed to calculate states and distances. No issues there.

However, when it comes to the density plot, things get a bit tricky. The problem arises, for instance, when dealing with the structure:

if structure:
    if structure.lattice == vol.lattice:
        plot_points(structure.cart_coords, structure.labels, fig=fig)
    else:
        plot_structure(structure, fig=fig)

This leads to plotting different lattices for the trajectory and the crystallographic positions extracted from the .cif file. Ideally, I'd like to visualize my lithium density around positions on the same lattice, but I find myself manually adjusting the lattice for the .cif structure to match my trajectory.

A similar problem appears during shape analysis. There's a warning, warn_lattice_not_close(self.lattice, test_lattice), but when it comes to calculating distances and cart_coords, self.lattice is the one in play. This can affect the size of sites that I'm looking at.

I think, users shouldn't be forced into lattice matchmaking. Therefore, we couldn't make it so that the lattices must be the same.

[stefsmeets]

Thanks for the clarification, would this help for plotting? Meaning you can specify your custom lattice to override the vol/structure lattices?

plots.density(
    vol = volume,
    structure = structure,
    force_lattice = volume.lattice,
)

[AstyLavrinenko]

Yes, I think it's perfect. And if someone wants to check visually the mismatch, structure.lattice could be forced

[stefsmeets]

I have two suggestions, either make the sites, lattice and space group directly configurable:

unique_sites = [sites[0] for sites in symmetrized_structure.equivalent_sites]

sa = ShapeAnalyzer(
    sites=sites, 
    lattice=trajectory.lattice,
    spacegroup=symmetrized_structure.spacegroup,
)
sa.analyze_trajectory(trajectory)

Or pass the lattice and drag it along:

sa = ShapeAnalyzer.from_structure(structure)
sa.analyze_trajectory(trajectory, force_lattice=trajectory.lattice)

I think the first has my preference, the second is going to be really ugly to program, and conceptually does not fit with the class structure.

[AstyLavrinenko]

I think we can go with the first option. And using ShapeAnalyzer.from_structure(structure) without forcing the lattice could be optional.