GroupOps.epsilon_factor_without_centering([0, 0, 0]) is twice Sgtbx result for centric space groups

[kmdalton]

I am unsure if this is a bug in Gemmi or a bug in Sgtbx. Either way it is probably academic.

The following script can be used to confirm that Gemmi sometimes calculates an epsilon factor without centering which is twice the value given by Sgtbx. Oddly, this only applies to reflection 0,0,0 and affects all centric space groups.

#!/usr/bin/env cctbx.python
from cctbx import sgtbx
import gemmi

h = [0, 0, 0]


for s in list(sgtbx.space_group_symbol_iterator()):
    xhm  = s.universal_hermann_mauguin()
    sg   = sgtbx.space_group(s)
    sgtbx_epsilon = sg.epsilon(h)
    gemmi_go = gemmi.SpaceGroup(xhm).operations()
    gemmi_epsilon = gemmi_go.epsilon_factor_without_centering(h)
    if gemmi_epsilon != sgtbx_epsilon:
        centric = gemmi_go.is_centric()
        print(f"{xhm:>12}\t{centric}\t{gemmi_epsilon}\t{sgtbx_epsilon}")