ENH: make use_bqstorage_api True by default

[tswast]

It’ll give people the best experience if the fast method is also the default method. Since this requires an additional API to be enabled, it’s technically a breaking change. Similar to how we handled the switch to standard SQL in googleapis/python-bigquery-pandas#195, I propose we complete this work in two phases:

• [ ] Add use_bqstorage_api to the global context object with a value of False. It is overridable by setting pandas_gbq.context.use_bqstorage_api = True. If the default value of use_bqstorage_api=None is passed into read_gbq, it uses the default value from the global context.
• [ ] Make a release with this new context value. (Make sure pandas.read_gbq also sends down None for use_bqstorage_api if not supplied.)
• [ ] Update default value in the global context to True. (Possibly after the BigQuery Storage API goes GA, though maybe sooner.)

[ketozhang]

Any news on when this will be implemented?

[tswast]

Nothing blocking it, now that the API is GA. We probably want to use the create_bqstorage_client parameter in the to_dataframe method for future compatibility.

[tswast]

This is long overdue. Bumping the priority on this.

[Et9797]

Doesn't use_bqstorage_api=True incur more cost? Or is kind of neglible when you're talking roughly 1-10GBs of data?

[tswast]

It's free to read from anonymous tables created by a query.