Cannot select atoms with StringSelection from empty AtomGroup

[lilyminium]

Expected behavior

empty_atomgroup.select_atoms('name anything') returns an empty AtomGroup.

Actual behavior

A TypeError is raised.

Code to reproduce the behavior

In [1]: import MDAnalysis as mda

In [2]: from MDAnalysis.tests.datafiles import PSF

In [3]: mda.__version__
Out[3]: '2.0.0-dev0'

In [4]: u = mda.Universe(PSF)

In [5]: ag = u.select_atoms('resname blahblah')

In [6]: len(ag)
Out[6]: 0

In [7]: ag.select_atoms('name O')
---------------------------------------------------------------------------
TypeError                                 Traceback (most recent call last)
<ipython-input-7-c43a12bdaf29> in <module>
----> 1 ag.select_atoms('name O')

~/pydev/mdanalysis/package/MDAnalysis/core/groups.py in select_atoms(self, sel, *othersel, **selgroups)
   2861             # Apply the first selection and sum to it
   2862             atomgrp = sum([sel.apply(self) for sel in selections[1:]],
-> 2863                           selections[0].apply(self))
   2864         return atomgrp
   2865

~/pydev/mdanalysis/package/MDAnalysis/core/selection.py in apply(self, group)
    517         for val in self.values:
    518             values = getattr(group, self.field)
--> 519             mask |= [fnmatch.fnmatch(x, val) for x in values]
    520         return group[mask].unique
    521

TypeError: ufunc 'bitwise_or' not supported for the input types, and the inputs could not be safely coerced to any supported types according to the casting rule ''safe''

Current version of MDAnalysis

• Which version are you using? (run python -c "import MDAnalysis as mda; print(mda.__version__)") 2.0.0-dev0
• Which version of Python (python -V)? 3.7
• Which operating system? MacOS

[RMeli]

I agree that the TypeError is really obscure, but wouldn't be better to have the selection on an empty atom group to fail cleanly instead of returning an empty atom group?

[richardjgowers]

I think it’s fairly common to chain selections, so an empty AG should return another empty AG. I can’t remember, is there a way to go to universe again via global or something?

…

On Wed, Jun 17, 2020 at 15:40, Rocco Meli ***@***.***> wrote: I agree that the TypeError is really obscure, but wouldn't be better to have the selection on an empty atom group to fail cleanly instead of returning an empty atom group? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#2765 (comment)>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACGSGBYA6G65HN46XP5MB73RXDIVJANCNFSM4OAGMC6Q> .

[lilyminium]

Yeah other ways to "select" atoms from empty AtomGroups also return an empty AtomGroup, so having select_atoms do the same seems more consistent.

>>> import MDAnalysis as mda
>>> from MDAnalysis.tests.datafiles import PSF
>>> u = mda.Universe(PSF)
>>> u.atoms[[]][[]]
<AtomGroup with 0 atoms>

I can’t remember, is there a way to go to universe
again via global or something?

I think group is an AtomGroup so you could try group.universe, if that's what you're asking. Anyway it looks like it's just the StringSelection with this issue and the below seems to fix it, although it may be superceded by #2755?

--- mask |= [fnmatch.fnmatch(x, val) for x in values]
+++ mask |= np.asarray([fnmatch.fnmatch(x, val) for x in values], 
+++                    dtype=np.bool)

Edit: I was wrong, the sphzone/layer selections have problems with reshaping a lack of center of geometry

[orbeckst]

Is this a fix that should be backported #2798 or is it a Change that can only go into 2.0.0?

[orbeckst]

Looking at PR #2767 you seem to be considering it a fix so I'll add it to the backport #2798. This will also add a CHANGELOG entry.