BQ: Empty array when no results

[hraban]

Better than KeyError

[theacodes]

I would prefer None over an empty list. How does json.dumps() respond when passed None?

[hraban]

$ python3 -c 'import json ; print(json.dumps(None))'
null

I went with [] because the return value is an array containing all the result rows, which feels like it should be an empty array when there are no rows. But I am almost apathetic towards this :P as long as it doesn't KeyError, I'm fine.

[theacodes]

Heh, fair enough... How about we do even better and do this:

rows = page.get('rows')

if rows:
    print json.dumps(rows)
else:
    print 'No results found.'

[hraban]

Not a fan. The appeal of this script is its composability through valid JSON output on stdout; you can pipe it to jq or other JSON tools. Incidentally this is also why I chose []; by always returning an array you can just loop over the result no matter what.

[theacodes]

Interesting, I don't think we considered that a primary use case here. How about:

    # [START paging]
    # Page through the result set and print all results.
    results = []
    page_token = None

    while True:
        page = bigquery.jobs().getQueryResults(
            pageToken=page_token,
            **query_job['jobReference']).execute(num_retries=2)

        results.extend(page.get('rows', []))

        page_token = page.get('pageToken')
        if not page_token:
            break

    print(json.dumps(results))
    # [END paging]

[hraban]

Ah! so you're saying paging would break the original script's JSON output?

This is a nice solution, good idea.

The valid JSON output is very nice in combination with jq because the raw output can be a bit.. raw :)

+1 !

[theacodes]

Cool, if you'll update your PR I'll get it merged. Thank you for your contribution!

[hraban]

Done.

[theacodes]

LGTM, merging. Thank you for your contribution!