AttributeError: AtomGroup has no attribute charges

[laurapulido]

I'm trying to use the potential_profile.py code available in the examples folder. After loading the topology and trajectory, I want to get the charges of all atoms in the system using the command universe.atoms.charges, but I get the following error:

AttributeError                            Traceback (most recent call last)
<ipython-input-26-a04736538ca1> in <module>
----> 1 universe.atoms.charges

/Applications/miniconda3/envs/MDAnalysis_Mdtraj/lib/python3.7/site-packages/MDAnalysis/core/groups.py in __getattr__(self, attr)
   2276                 pass
   2277         raise AttributeError("{cls} has no attribute {attr}".format(
-> 2278             cls=self.__class__.__name__, attr=attr))
   2279 
   2280     def __reduce__(self):

AttributeError: AtomGroup has no attribute charges
Actual behavior

What happened instead. Add as much detail as you can. Include (copy and paste) stack traces and any output.

Thank you for your help

[orbeckst]

Does your topology file contain charges? The .charges attribute is only created if you have charges.

How did you load your Universe, i.e., what input files do you have?

[richardjgowers]

So on a dev level, maybe this message should grab some debug info from Universe (filenames) to make the message clearer?

[zemanj]

So on a dev level, maybe this message should grab some debug info from Universe (filenames) to make the message clearer?

Yeah that will save some headaches. 👍

[zemanj]

@richardjgowers BTW what is the rationale behind this whole AtomGroup argument injection machinery besides uncluttering tab completion?
IMHO raising exceptions due to missing TopologyAttrs should be the responsibility of Topology.
The fact that missing transplanted properties/methods raise unhelpful AttributeErrors has the potential to confuse users and has already led to some dirty workarounds on the dev side, see e.g. in *Group.wrap() and *Group.unwrap() (and there's more of that kind elsewhere).

[richardjgowers]

Yeah mostly to try and not clutter AtomGroup will every attribute and method we support across all formats. Maybe we can raise a different error (NoDataError), but I’m slightly in favour of the error here. Here’s an example of where we catch an error and do better: https://github.com/MDAnalysis/mdanalysis/blob/develop/package/MDAnalysis/core/groups.py#L2218 All that said, maybe we should include more attributes by default, ie charge, mass, name. I’ve been thinking of making AtomGroup a cython class, which ties in with this...

…

On Tue, Feb 18, 2020 at 08:52, Johannes Zeman ***@***.***> wrote: @richardjgowers <https://github.com/richardjgowers> BTW what is the rationale behind this whole AtomGroup argument injection machinery besides uncluttering tab completion? IMHO raising exceptions due to missing TopologyAttrs should be the responsibility of Topology. The fact that missing transplanted properties/methods raise unhelpful AttributeErrors has the potential to confuse users and has already led to some dirty workarounds on the dev side, see e.g. in AtomGroup.wrap() <https://github.com/MDAnalysis/mdanalysis/blob/1e3f3627990353cf0d5ee00a567e563f012de624/package/MDAnalysis/core/groups.py#L1440> and AtomGroup.unwrap() <https://github.com/MDAnalysis/mdanalysis/blob/1e3f3627990353cf0d5ee00a567e563f012de624/package/MDAnalysis/core/groups.py#L1584> (and there's more of that kind elsewhere). — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#2536?email_source=notifications&email_token=ACGSGB53GHZLNBMIOKR64M3RDOOUZA5CNFSM4KW3SMXKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEMBDONI#issuecomment-587347765>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/ACGSGB6R2DL3HDERV2W2M7DRDOOUZANCNFSM4KW3SMXA> .

[zemanj]

Here’s an example of where we catch an error and do better: https://github.com/MDAnalysis/mdanalysis/blob/develop/package/MDAnalysis/core/groups.py#L2218

I see your point, but keep in mind that this solution means we have to do the bookkeeping manually.
Also, the error mesage "<attr> not available; this requires Bonds" could be improved, especially because the error is often not triggered by the user trying to directly access some AtomGroup attribute but rather when trying to run some method such as unwrap().
Maybe the error message could be generalized to something like "The attribute/method you are trying to access is not available because your topology <insert topology filename if possible> doesn't contain information about <insert corresponding TopologyAttr>."

All that said, maybe we should include more attributes by default, ie charge, mass, name. I’ve been thinking of making AtomGroup a cython class, which ties in with this...

I'm all in favor of this. However, if we have attributes like masses and charges by default, shouldn't that also apply to related methods such as center_of_mass(), total_charge(), etc?

[richardjgowers]

Yeah so associated methods would get included in with the mandatory attributes. One thing that would get a lot faster would be AG.center_of_mass() because it would be all C(++).

[zemanj]

@laurapulido To give you a quick answer: The most likely reason for the error you are seeing is that the topology file format you used to construct your Universe doesn't contain charge information.
Unfortunately, the error message is not very clear about this.

I would appreciate it if you could tell us whether my (and @orbeckst's) guess is right (or wrong).

[laurapulido]

Thank you for all your answers. I checked the topology file and I found that it doesn't have charges.

Thank you

[zemanj]

Thanks @laurapulido for reporting back!
@richardjgowers @orbeckst Should we close this and open a new issue regarding improving the error handling/message?

[orbeckst]

sure

…

On Feb 19, 2020, at 5:25 PM, Johannes Zeman ***@***.***> wrote: Thanks @laurapulido <https://github.com/laurapulido> for reporting back! @richardjgowers <https://github.com/richardjgowers> @orbeckst <https://github.com/orbeckst> Should we close this and open a new issue regarding improving the error handling/message? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#2536?email_source=notifications&email_token=AAB2DHDEFMOSHDNAOGRE7HDRDXEXXA5CNFSM4KW3SMXKYY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOEMKH4OA#issuecomment-588545592>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AAB2DHHR5WTCTRLVBZEMQKTRDXEXXANCNFSM4KW3SMXA>.

-- Oliver Beckstein orbeckst@gmail.com * orbeckst@gmx.net

[orbeckst]

@zemanj – you have been put in charge for closing this issue and opening a new one :-)

[zemanj]

Superseded by #2565, closing.