Add a config option to disable caching?

[adrn]

Via email, @dnidever asked me if there was a way to disable caching with astroquery. It looks like each .query*() method has a cache=False option, but is there a way to do this globally? It would be nice to add a config option to handle this, if it doesn't already exist.

[dnidever]

Thanks for opening the issue Adrian.
I'm using Vizier.query_region() and don't see a cache=False option. I also don't see any Vizier.cache attribute only cache_location.

[keflavich]

Right, <astroquery module>.cache_location is the directory it writes to; a hacky workaround is to do Vizier.cache_location = '/dev/null'. There should be cache=T/F options for every module....

[keflavich]

query_region accepts the cache=<bool> option, but it's not documented properly.

[eteq]

@keflavich - isn't it also possible to have cache_location be None? I know there's at least some code that checks for that (although I'm not sure how uniformly implemented it is...)

[eteq]

(If so, I suppose it should be documented clearly, though!)

[keflavich]

Huh, yes, that's correct:
https://github.com/astropy/astroquery/blob/master/astroquery/query.py#L235

[dnidever]

Thanks. I'll try out the cache_location=None option. But that seems like a non-intuitive way of turning off caching to me.

[keflavich]

Can you suggest what would be most intuitive? I agree that it's not obvious, since it's not documented at all, but that would probably have been my first suggested approach once we do document it.

[dnidever]

I find the cache=True/False option intuitive and it would be consistent with the other query routines. Also, I wonder if the default option should be False. The caching filled up my home directory and it took me quite a while to figure out what was going on.

…

On Thu, Jul 25, 2019 at 5:38 PM Adam Ginsburg ***@***.***> wrote: Can you suggest what would be most intuitive? I agree that it's not obvious, since it's not documented at all, but that would probably have been my first suggested approach once we do document it. — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#1511?email_source=notifications&email_token=ACHU3N2EPWTA2RWVVKCK2MDQBI2OVA5CNFSM4IGTNCK2YY3PNVWWK3TUL52HS4DFVREXG43VMVBW63LNMVXHJKTDN5WW2ZLOORPWSZGOD23CIJQ#issuecomment-515253286>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ACHU3N2OGSL3XIXLKYKFITLQBI2OVANCNFSM4IGTNCKQ> .

-- Dr. David Lee Nidever Assistant Professor Department of Physics Montana State University Barnard Hall (EPS building), Room 202 Bozeman, MT 59717

[keflavich]

OK, that sounds good - cache=True/False exists and is mostly documented, but some documents haven't propagated (because of the way _async vs sync functions are defined).

I'm torn about defaulting caching to False, since that can make a lot of potentially cheap operations very expensive. But, I think caching in its present form isn't well enough communicated, as you point out. Some options we should consider:

(1) Warn the first time caching is done in a session so users know that caching happens
(2) Cache to tempfiles by default so that they automatically clear at the end of session. This prevents caching between sessions, but maybe that's just a generally better approach anyway?

Any other ideas? Neither of these is difficult to implement, and I think they're both reasonable.

[bsipocz]

#1634 meant to address this, reopen if it's not the case.