question about SpaceGroup GroupOp order about gemmi HOT 8 CLOSED

I don't know. If operations in Jeremy Cockcroft's tables or in International Tables of Crystallography are ordered according to some rules, I could try to use these rules in gemmi.
But perhaps the order is just random, I don't know.

from gemmi.

hmm the obvious answer would be that the order is determined by the successive application of the symmetry generators ; in this case 2_1 = is a 180° (twofold) rotation followed by a translation of 1/2 of the lattice vector, so the second entry must in fact be -x+1/2,y+1/2,-z+1/2 in this case.

In the documentation you state that gemmi derives its symmetry operators from the HM symbols, so the order should be correct when generating? - or is there a sorting step in between

from gemmi.

Operations are derived from Hall symbols. H-M P 1 21/n 1 corresponds to Hall symbol -P 2yn,
which encodes two generators:
-x,-y,-z
-x+1/2,y+1/2,-z+1/2
The first one is encoded by -. If I'd put the -x,-y,-z generator as the last one I'd get the same order for this space group.
This is a trivial change that I can do. But I don't know if helps with other space groups.

from gemmi.

I did this change. For P 1 21/n 1 the order is now the same as on the linked page. I didn't check other space groups.

from gemmi.

thanks, but I would ask you to revert the change. I have tested with spglib, which gives the same order as gemmi (before modification). I think I found that the spacegroups are decomposed into the cosets under the translation group ... I need to figure out exactly how to do this, maybe later I can provide a function for getting the ITA order.

from gemmi.

But on the other hand the new order is the same as in cctbx and CCP4 syminfo.lib.

from gemmi.

ok. Have you tested for other space groups as well?
I also suspect is is just a question of how the data is stored (triplet form for all or generated)
but the symmetry operations order must be defined somehow (the cif field _symmetry_equiv_pos_as_xyz should be the same always no matter which program generated the cif file). .. if I find out something I will report back. Thanks

from gemmi.

I don't think that the order matters. The coreCIF spec only requires that x,y,z has id=1. Apart from this the order of _symmetry_equiv_pos_as_xyz is not specified.

from gemmi.