`universe.<segment id>` does not issue a deprecation warning

[jbarnoud]

Expected behaviour

When using a segment instant selector from a Universe instance, a deprecation warning is issued. Indeed, instant selectors are deprecated and #1403 introduced deprecation warnings for them.

Actual behaviour

No deprecation warning is issued.

Segment instant selector are added directly to the universes name space here: https://github.com/MDAnalysis/mdanalysis/blob/develop/package/MDAnalysis/core/universe.py#L345. Without a proxy, accessing the segments cannot issue the warning.

This is already acknowledge in a comment: https://github.com/MDAnalysis/mdanalysis/blob/develop/package/MDAnalysis/core/universe.py#L322. However, this is the most documented way to use instant selectors.

Code to reproduce the behaviour

import MDAnalysis as mda
from MDAnalysisTests.datafiles import PSF, DCD

u = mda.Universe(PSF, DCD)
u.s4AKE

[richardjgowers]

You really do find all the best bugs -_-.

…

On Thu, 13 Jul 2017, 12:55 p.m. Jonathan Barnoud, ***@***.***> wrote: Expected behaviour When using a segment instant selector from a Universe instance, a deprecation warning is issued. Indeed, instant selectors are deprecated and #1403 <#1403> introduced deprecation warnings for them. Actual behaviour No deprecation warning is issued. Segment instant selector are added directly to the universes name space here: https://github.com/MDAnalysis/mdanalysis/blob/develop/package/MDAnalysis/core/universe.py#L345. Without a proxy, accessing the segments cannot issue the warning. This is already acknowledge in a comment: https://github.com/MDAnalysis/mdanalysis/blob/develop/package/MDAnalysis/core/universe.py#L322. However, this is the most documented way to use instant selectors. Code to reproduce the behaviour import MDAnalysis as mdafrom MDAnalysisTests.datafiles import PSF, DCD u = mda.Universe(PSF, DCD) u.s4AKE — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#1478>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AI0jB6m64c2iT833tyjMlZ4YJzrmfy0wks5sNgWsgaJpZM4OW4uu> .

[richardjgowers]

Ok thinking about this, it might be possible to make u.SEGNAME actually a property, which first issues the dep warning and then returns the segment. So a little shim that makes this work correctly.

[jbarnoud]

On 07/13/2017 03:23 PM, Richard Gowers wrote: Ok thinking about this, it might be possible to make |u.SEGNAME| actually a property, which first issues the dep warning and then returns the segment. So a little shim that makes this work correctly. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#1478 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABUWuoms2xz1uf16beiAk9Kv_MyhYhJZks5sNhpdgaJpZM4OW4uu>.

I had a brief look at descriptors, but I did not go through. The issue I see is that descriptors (and therefore properties that are just a shortcut to the descriptor mechanism) are attached to the class and not to the instance. I am not sure we can have instance wide descriptors.

[richardjgowers]

The other way would be to make them work like ag.CA (via a hack in __getattr__), but this would then lose the tab completion.. which is probably fine tbh.

[jbarnoud]

On 07/13/2017 04:04 PM, Richard Gowers wrote: The other way would be to make them work like |ag.CA| (via a hack in |__getattr__|), but this would then lose the tab completion.. which is probably fine tbh. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#1478 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABUWusu42y-gnp0r2mGyYM6-Yxel-YZuks5sNiP_gaJpZM4OW4uu>.

Would we loose all tab-completion or only the tab-completion of the segment names? In the former case it would be annoying, in the latter it is not be a big deal, the feature is deprecated anyway.

[richardjgowers]

the latter, the segments wouldn't exist in the Universe namespace for tab completion inspection to see.

[jbarnoud]

On 07/13/2017 04:10 PM, Richard Gowers wrote: the latter, the segments wouldn't exist in the Universe namespace for tab completion inspection to see. — You are receiving this because you authored the thread. Reply to this email directly, view it on GitHub <#1478 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABUWuudCatWk6NiKuig94gVcwqgfKHtEks5sNiVUgaJpZM4OW4uu>.

Then it is probably the better solution. Loosing tab-completion for a feature we are phasing out is not really a problem in point of view.