Weird behaviour in multi_component_hydrogen_bond_propensity_report.py

[MilitaoLucas]

When using the multi_component_hydrogen_bond_propensity_report.py I found that

ascores[a.label] = [round((coordination_scores.predictions_for_label(a.label, 'a')[1])[k], 3)

raises 'NoneType' object is not subscriptable
It may look like some problem with the script, but on a deeper investigation,

coordination_scores.predictions_for_label(a.label, 'a')

returns None type. The label in question is O1_B.
Investigating a little I discovered that in coordination_scores._predictions the prediction with O1_B was present.

(('O1_B of carboxylate (a)', 0),
                                        ('0.146488', 0),
                                        ('0.577361', 1),
                                        ('0.259616', 0),
                                        ('0.0149453', 0),
                                        ('0.0015899', 0)),

It seems like

if not self._predictions:
    return None

In the CrystalDescriptors Class is somehow returning None, as the other return should return a Tuple, even if it is empty. I don't know the type of the coordination_scores._predictions tuple like object, but if it implemented an __bool__ like function to define a different default behaviour, it could be the source of this problem.
I am having this problem specifically doing the multi_component_HBP of FUTWOI37 and FOXLUA. After updating to the 3.1.0, the problem that R couldn't run returned (installed with conda). I also found the same problem in windows (installed with pip). So, the fix I was using could be the source of this problem even if I don't see how. Furthermore, it works fine for every other pair of FOXLUA-->Coformer but not FUTWOI37. So, I would like to know if you can reproduce this potential bug. Investigating the code of the API that isn't compiled, I could only come to the conclusions presented above.

[pmbulit]

Hi @Luke2Sky , both refcodes you mention have disorder present, which makes things difficult to handle downstream. Are you running the script with CSD refcodes as input or with mol2/cif files? If it is files, could you forward them to support@ccdc.cam.ac.uk for me to have a look?

[MilitaoLucas]

I am running with mol2 files from CSD. I think better error handling on the API side would be appreciated. I will forward the files. Should I put a code or something in the subject to better refer to this issue? Maybe the issue ID? I will see if I can find entries with less disorder. Maybe a warning would be fine. Also, I will soon submit the patch I made so that multi-component uses try excepts to avoid stopping the entire run when this problem happens with one coformer.

[pmbulit]

You can add the issue number on the subject of the email.

[MilitaoLucas]

I just found that COD has CaCO3 without disorder. I will test every case were structures have disorder before mailing support. The mol2's where obtained using EntryReader and CrystalWriter.