read_gbq should offer an option to save to dataset

[iros]

Right now read_gbq runs a query and retrieves the results into a data_frame. If one wants to also save it into a dataset table, one then has to call to_gbq which would stream the results back.

The bq command line tool has an option to pass a destination dataset table immediately on query to store the results.

cat queries/sample.sql | bq query --format csv 
  --destination_table=collection.table_name > output.csv

Ideally read_gbq would have an additional option like destination_table that could store the results immediately.

[jorisvandenbossche]

cc @parthea @aaront @jacobschaer

[iros]

It looks like the configuration obj for the query job here can take a destinationTable parameter, so we just have to carry it through:

https://github.com/pydata/pandas/blob/f71537ab2561ab5727008095e9685966619fa7b9/pandas/io/gbq.py#L325-L331

[jreback]

xref to #10474 not really sure we should support this

[parthea]

I'm ok with closing this feature request. I'll keep PR #11209 in my back pocket in case this gets raised again.

Looking at #11209 again, it would be cleaner to create a new api function gbq.query_to_table() rather than re-define the existing read_gbq() which returns a DataFrame

[parthea]

@jreback Can we close this as duplicate of #10474 , and continue the discussion there?

[jreback]

sure