GeezerVison

If you tried to push an item which contained a NaN to a dataset, it would return an error from the Apify API, that the JSON you sent to it is invalid.

>>> client.dataset('~nan-testing-dataset').push_items([{ 'a': float('NaN')}])
ApifyApiError: Cannot parse POST payload as JSON: Unexpected token N in JSON at position 7

That is because in Python, by default, json.dumps() stringifies NaN and Infinity values. This is against the JSON standard, and JSON.parse on the API fails to parse the value.

This fixes it by adding the allow_nan=False parameter to json.dumps(), so that the error is caught before sending the invalid JSON to the API, and throws an error early.

>>> client.dataset('~nan-testing-dataset').push_items([{ 'a': float('NaN')}])
ValueError: Out of range float values are not JSON compliant: nan

By the way, when fixing this, I made my first fix to CPython: python/cpython#99926