Use google-auth library and requests instead of google-api-client

[tswast]

The google-auth library provides a Requests interface.

The google-auth library is preferred over the oauth2client library.

Would this project be amenable to a PR to update how the requests to BigQuery are made?

[max-sixty]

My working hypothesis is we should transition to the google.cloud.bigquery library over time, and that library can wrap any auth calls.

But I'm not a maintainer, and I may be missing something - is there a reason we would need to auth separately from that library?

[tswast]

Yeah google.cloud.bigquery would be okay, but since we only need the query feature it might be overkill. Also, google-auth is 1.0, but google.cloud.bigquery is still going through breaking changes.

[jreback]

how much work would be involved to accept this an alternative? iow you try to import oauth2client if that fails you can this package else give up

[parthea]

I'm happy to help with testing/code review for this addition.

[parthea]

how much work would be involved to accept this an alternative? iow you try to import oauth2client if that fails you can this package else give up

This sounds great ( or vise-versa, if google-auth , else if oauth2client else give up)

[tswast]

What's the reason for accepting either library?

[parthea]

What's the reason for accepting either library?

@tswast I see your point. Assuming that we can maintain the existing functionality of pandas-gbq with google-auth, then I agree we could drop oauth2client. @jreback , would that be ok?

Note: Historically, oauth2client was an optional dependency for pandas so we had to deal with the scenario where users did not have certain libraries. Now that we have migrated pandas.io.gbq to pandas-gbq as a separate package from pandas we have more control over the dependencies.

[jreback]

ok with changing dependencies. Though nice to have some back-compat. IOW I ok just adding this library, then deprecating the old.

[max-sixty]

You guys know this better, but is it worth considering google.cloud.bigquery?

We have built a set of internal functions on this (which we're planning to move into this library). It takes care of a lot of the boilerplate.

Is there a compelling reason not to add this as a dependency?

[jreback]

thats #25; again which am happy to take, assuming things can be built/installed/tested properly.

[max-sixty]

thats #25; again which am happy to take, assuming things can be built/installed/tested properly.

Then I think we can also use that library for auth (I believe it wraps google-auth), and this discussion would be moot.

I commented here; we can still use json keys with either the oauth2client or google-auth if we want to enable that

[tswast]

I'll close this out in favor of eventually moving the API-calling code to google-cloud-python. I've got a separate PR open to move to the new auth library without changing the client libraries.

The queries wouldn't be too bad to do with a pure REST call, but the file uploads get a bit hairy, so having a library help with that is ideal.