Cannot insert NaN and non-finite float values using Client.insert_rows()

[wmdrichards]

Trying to insert a NaN using Client.insert_rows fails. I tracked it down to

def _float_to_json(value):
    """Coerce 'value' to an JSON-compatible representation."""
    return value if value is None else float(value)

which should return "NaN" to be compatible with the REST API. Currently it returns an actual nan float value. Probably similar issues occur with other non-finite floats.

Environment details

• OS type and version:
• Python version: Python 3.6.8
• pip version: pip 21.1
• google-cloud-bigquery version: 2.20.0

Steps to reproduce

1. Try to insert a NaN value into existing table

Code example

from google.cloud.bigquery import Client

client = Client(project='bar')
client.insert_rows(table=client.get_table('foo'), rows=[{'bar': float('NaN')}])

Stack trace

google.api_core.exceptions.BadRequest: 400 POST (REDACTED): Invalid JSON payload received. Unexpected token.
json": {"bar": NaN}, "insertId": "1

[wmdrichards]

For anyone else running into this, using insert_rows_json is a valid workaround, as long as the row is already json-compatible

[plamut]

Thanks for reporting and researching that. It turned out that non-finite floats were indeed problematic, too.