Define path of bigquery_credentials.dat by environment var

[jonathansp]

In a shared environment, the use of hard-coded bigquery_credentials.dat file name could be a problem. We should be able to define this path by using an environment variable, for instance, PANDAS_GBQ_CREDENTIALS_FILE.

[max-sixty]

Can we default to GOOGLE_APPLICATION_CREDENTIALS, or does this need to be additional? https://developers.google.com/identity/protocols/application-default-credentials

[parthea]

Thanks for reporting this. I think this is a duplicate of #41. We have an improvement ready (#83) and I'm hoping to merge it tonight if there are no objections. Could you add comments to #83?

[andrelu]

IMO it would be interesting to have a configurable path. In #83 we cannot set a default directory to store our config, you ll always store in ~/.config/pandas_gbq... this is not interesting when having datascience isolate notebooks environments. Beside that, you always have to pass the credentials when executing read_gbq.

[andrelu]

@MaximilianR I think that GOOGLE_APPLICATION_CREDENTIALS env is used for a different kind of credential.

[jonathansp]

Thank you guys for replying.

@parthea @MaximilianR could you please consider both approaches? Maybe it'll be tough asking our users for sending credentials_path parameter every time they need to exec a query.

@MaximilianR, as @andrelu pointed out, GOOGLE_APPLICATION_CREDENTIALS seems to have another meaning.

[parthea]

@jonathansp You're right. Users may not like sending credentials_path parameter every time they need to exec a query. I'll modify #83 tonight to use an environment variable, as you've done in #86, and remove credentials_path altogether. I'll start by building on top of your changes in #86 so we can keep that PR open to maintain authorship.

[max-sixty]

I think that GOOGLE_APPLICATION_CREDENTIALS env is used for a different kind of credential.

Yes, you're right

[parthea]

Closed via dbfb4e9