BigQuery: preview / list_rows command for %%bigquery Jupyter magic

[tswast]

Is your feature request related to a problem? Please describe.

When you want to get all rows in a table (or just a preview), it's more expensive to run a SELECT * query than the alternatives.

Describe the solution you'd like

• If there is no whitespace in the query text, assume it's a table ID. Call list_rows (possibly with the BigQuery Storage API option, if that's set).
• To allow for a preview, add a --max_results option to the %%bigquery magic. If not set, get all rows. If set, download at most max_results rows.

Describe alternatives you've considered

• Add a special %%bigquery preview command. I prefer looking to see if the query text is just a table name, as that will be more consistent with pandas-gbq. Also, we'd prefer to limit the %%bigquery magic to just "queries". I think entering just a table ID as query text is clear that it's different from SELECT * (anti-pattern) but also the intention is clear.
• Client-side query processing to automatically detect SELECT * [LIMIT N] queries client-side. I think it could make sense to parse queries client-side in just the %%bigquery magic (or opt-in via QueryJobConfig option), but this is much more complex than checking if there is any whitespace in query text.

Additional context

Related feature requests in pandas-gbq:

• List rows without a query. ENH: Read BigQuery query table without a query python-bigquery-pandas#266
• Limit rows returned. Add option for limiting rows of retrieved of results python-bigquery-pandas#102

[tswast]

CC @shubha-rajan

If you're interested, I think this could be a natural next task to get more familiar with the BigQuery API(s) and the Jupyter features.

[shubha-rajan]

@tswast sure, I'll take this one!

[plamut]

With #9170 now merged, this issue can be closed.