Speeding up get_nn_info in local_env.py

[ftherrien]

_get_image and _get_original_site do not need to recompute site and index when site is a PeriodicNeighbor which has them as attributes.

When there are several sites, this makes a significant difference on speed for get_nn_info

Alternative implementation would have been try/except and hasattr. try/except is much slower when most of the sites are not PeriodicNeighbor.

Summary

Major changes:

• feature: image and site index are not recomputed when they are already an attribute of the site

[JaGeo]

Awesome. Thanks! I recently tried the minimum distance method for an amorphous structure 😅...
And it took forever.

[janosh]

@ftherrien thanks, much appreciated! the failing test here looks like maybe this doesn't handle translations correctly? shifting the I atom by a full $a$ lattice vector shouldn't make a difference.

TestCrystalNN.test_shifted_sites _______________________

self = <tests.analysis.test_local_env.TestCrystalNN testMethod=test_shifted_sites>

    def test_shifted_sites(self):
        cnn = CrystalNN()
    
        sites = [[0.0, 0.2, 0.2], [0, 0, 0]]
        struct = Structure([7, 0, 0, 0, 7, 0, 0, 0, 7], ["I"] * len(sites), sites)
        bonded_struct = cnn.get_bonded_structure(struct)
    
        sites_shifted = [[1.0, 0.2, 0.2], [0, 0, 0]]
        struct_shifted = Structure([7, 0, 0, 0, 7, 0, 0, 0, 7], ["I"] * len(sites_shifted), sites_shifted)
        bonded_struct_shifted = cnn.get_bonded_structure(struct_shifted)
    
>       assert len(bonded_struct.get_connected_sites(0)) == len(bonded_struct_shifted.get_connected_sites(0))
E       assert 1 == 2

[janosh]

thanks again @ftherrien!

[ftherrien]

@janosh the auto-merge was cancelled FYI

[JaGeo]

I have merged now, @janosh , as you approved it already.

[ftherrien]

Thanks y'all!