Does Dihedral analysis use dimensions from separate universes?

[orbeckst]

It looks as if DihedralAnalysis takes a short cut and only uses the dimensions of the first ensemble member:

MDPOW/mdpow/analysis/dihedral.py

Lines 102 to 103 in 2092cfd

	angle = calc_dihedrals(cord1, cord2, cord3, cord4,
	box=self.g1[key_list[0]].dimensions)

This can lead to wrong results. Instead, the calculations need to be done for each universe separately.

[orbeckst]

@ALescoulie can you please explain what's happening in DihedralAnalysis — did I misunderstand how it works?

[ALescoulie]

@orbeckst In _single_frame it's looking at stuff from a single universe, I just split in into 4 EnsembleAtomGroups. At the beginning _reorg_groups takes the given dihedral selections and makes 4 EnsembleAtomGroups each just containing the one atoms of each dihedral in the inputted dihedrals. This done for efficiency (I think I found that it was slower in my report) since we could do the dihedral calculations together with only one call to Cython. One of the functions called in __init__ ensures the passed in list is all from the same Ensemble. I don't think this is an issue.

[orbeckst]

Thank you for the explanation. I didn't understand how the current universe is selected. This all makes more sense now.

(Incidentally, you had actually fixed the box in #198 ).