Description
I've recently noticed that pandas_gbq
requires that a user's home directory exist and be writeable in order to be used. This means that it cannot be used in restricted environments, where that may not be the case.
I think this is a result of the recent change (#241) to use https://github.com/pydata/pydata-google-auth for auth.
I've raised the underlying issue as pydata/pydata-google-auth#10, however from what I can tell pydata_google_auth
is only actually used in the case where the user doesn't provide credentials as part of an operation.
In particular, I'm using a service account (where you're more likely to be in a restricted environment too), and always provide the credentials upfront anyway.
It would therefore be great if pydata-google-auth
was a soft requirement rather than a hard one -- both removing it from the requirements of this library and from the checks in _test_google_api_imports
, or making those checks aware of whether credentials have been explicitly provided.