migrate symmetry perception algo's away from Molecule?

[nickvandewiele]

Not an error but rather an 'idea' for rmg-py's architecture

I was currently revising the symmetry perception algorithms in rmg-py. They consist of several sub sections (atom, bond, axis, etc) as envisioned by Joanna Yu if I am not mistaken. In RMG-Java many of the molecule symmetry perception algorithms are slightly modified to internal rotors (as I am currently experiencing in trying to port those algorithm's to rmg-py).

Would you not consider removing all the perception algorithms from Molecule into a separate dedicated calculator type? Molecule is already getting quite heavy-weight (if you ask me) and symmetry perception is not really a core part. Moreover, then we could have a look how to unify both the rotor symmetry perception algo's and the regular molecule symmetry algo's into more generally applicable procedures.

Just my two cents... :)

[rwest]

I agree that Molecule is getting rather large (we noticed this as @pierrelb began adding geometry methods to it). My first impression is that this suggestion sounds ok, but @jwallen may have thought more about it.

[jwallen]

I think this is a good idea. Perhaps a new module rmgpy.symmetry? We could also make a whole rmgpy.molecule subpackage to encapsulate everything (symmetry, geometry, isomorphism, drawing, etc.) At any rate, looks like I haven't done much Cythonizing of those functions, so you can probably leave them as pure Python for the time being.

[goldmanm]

Symmetry has been migrated to rmgpy/molecule/summetry