Enhancement request: Ignore NaN

[colinalexander]

Similar to simplejson, would it be possible to include a parameter to ignore NaNs?

d = {0: -0.8285523717159029, 1: nan}

Current Behavior

>>> ujson.dumps(d)
---------------------------------------------------------------------------
OverflowError                             Traceback (most recent call last)
<ipython-input-88-77a2055a90c0> in <module>()
----> 1 ujson.dumps(d)

OverflowError: Invalid Nan value when encoding double

Desired Behavior

>>> simplejson.dumps(d, ignore_nan=True)
'{"0": -0.8285523717159029, "1": null}'

[jzwinck]

Pandas' version of ujson does just this. Try it:

import pandas
print(pandas.json.dumps({'foo': np.nan}))

{"foo":null}

You might be able to use it in the meantime.

[Jahaja]

This is a duplicate of #146, but I'm leaning towards in adding after all due to our ongoing effort to align more with the standard json module in v2.0.

[hexvolt]

in newer pandas versions:

pandas.io.json.dumps({'foo': np.nan})