make empty sphzone / cyzone selections consistent with 'around'

[xiki-tempula]

Is your feature request related to a problem?

Since sphzone will do a center of geometry calculation first, if an empty atomgorup is passed to the sphzone, a not very useful error message will be given.

Traceback (most recent call last):
    water = u.select_atoms( 'resname SOL and sphzone 3 (resid 12 and name HE)')
  File "/Users/grte2001/GitHub/mdanalysis/package/MDAnalysis/core/universe.py", line 677, in select_atoms
    return self.atoms.select_atoms(*args, **kwargs)
  File "/Users/grte2001/GitHub/mdanalysis/package/MDAnalysis/core/groups.py", line 2863, in select_atoms
    selections[0].apply(self))
  File "/Users/grte2001/GitHub/mdanalysis/package/MDAnalysis/core/selection.py", line 136, in apply
    rsel = self.rsel.apply(group)
  File "/Users/grte2001/GitHub/mdanalysis/package/MDAnalysis/core/selection.py", line 168, in apply
    return func(self, group)
  File "/Users/grte2001/GitHub/mdanalysis/package/MDAnalysis/core/selection.py", line 333, in apply
    ref = sel.center_of_geometry().reshape(1, 3).astype(np.float32)
ValueError: cannot reshape array of size 0 into shape (1,3)

Describe the solution you'd like

An error message which says the selection is empty could be given.

EDIT: Following the discussion below and on #3026 we want these selections to behave like around, i.e., just return an empty AtomGroup. — @orbeckst

Describe alternatives you've considered

Just return an empty atomgroup.

[IAlibay]

That's odd, I can't seem to reproduce this locally with:

u = mda.Universe(GRO, XTC)
u.select_atoms('resname SOL and sphzone 3 (resid 12 and name HA)')

[xiki-tempula]

@IAlibay Sorry, I noticed that I have made an error and this is a different problem.

[IAlibay]

So the behaviour here is inconsistent between around & sphzone/cyzone. In the former if the selection is an empty atomgroup we just return an empty atomgroup:

mdanalysis/package/MDAnalysis/core/selection.py

Lines 276 to 280 in 23074c6

	# All atoms in group that aren't in sel
	sys = group[~np.in1d(group.indices, sel.indices)]
	if not sys or not sel:
	return sys[[]]

whilst in the latter selections that check isn't being made.

My proposal here would be to align everything to around. IMHO it's a more commonly used selection keyword and the behaviour would more likely be the one that would be expected?

[orbeckst]

I agree with #2915 (comment) - make sphzone/cyzone behave in the same way as around and return and empty AtomGroup.

[orbeckst]

Note that some more details are shown in the duplicate issue #3026.

[sulays]

Hi, I was going through the code and found out that many other selections are also not consistent with around and they don't return an empty AtomGroup (e.g sphlayer). Should I create a PR for changes in all such selections?

[xiki-tempula]

Hi @sulays Thanks for your interest. I think this is a very good idea. It might be good to raise these selections as issues (explain what you expect and what you got) and then make a PR against them.