dials.cosym does not change space group to the requested one (small molecule case)

[dagewa]

Working on cRED data sets with @aimeon, we found a case where dials.cosym does a surprising thing: bad-cosym.zip

The integrated data all have space group $P2_1/m$, which actually we know to be correct. When cosym was run like this (ignoring the fact that there is no point in running cosym in this instance):

dials.cosym integrated_{1..5}.{expt,refl} space_group="P2_1/m"

The output symmetrized.expt had been changed to $P2_1$, despite the fact that we asked for $P2_1/m$. Is this a case of cosym not understanding the non-Sohncke groups? But even so, it should respect the requested space group, right?

[huwjenkins]

(ignoring the fact that there is no point in running cosym in this instance)

I think this is part of the (wider) issue though.

Your input datasets are indexed consistently and you know the space group so really you want dials.cosym to do nothing here. You could just pass integrated_{1..5}.{expt,refl} directly to dials.scale.

Aside from the space group change, by default with the command line you used dials.cosym will spuriously reindex 3 of the input datasets. That can be avoided by either setting lattice_group=P2 or lattice_symmetry_max_delta=1 but should the user setting space_group override more of the decision making in dials.cosym? That was also suggested by @biochem-fan here: #2320 (comment)

edit - make it clear(er) that it's the addition of the space_group parameter that leads to the reindexing.

[dagewa]

I didn't see 3 of the input datasets reindexed, but all 5 were reindexed in the same way on my run:

-x,y+z,-z: [0, 1, 2, 3, 4]

If I omit the space_group the it does no reindexing, but the space group is changed (from $P2_1/m$ in the input files to $P2$ in the output).

I feel I don't know how to use cosym. There is no PHIL help for the lattice_group or lattice_symmetry_max_delta parameters.