Numpy NaN handling

[zifeo]

Currently when a [Float] type resolves to np.array([1, np.nan]) the generated payload is [1, NaN] which is not json-compilant (rather than null) without any warning. A custom scalar can do the job however as numpy is so widely used, maybe this should be supported out of the box?

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[Manelmc]

This is still happening, is there any particular reason for resolving np.nan with NaN in type Float? I think a json compliant solution (null as @zifeo suggested) should be supported.

[zifeo]

Still an issue

[jkimbo]

@zifeo if you would like to create a PR with a custom scalar for Numpy I will happily merge it. Unfortunately I don't have the time to work on this at the moment.

[zifeo]

@jkimbo not sure when I will have the time yet, however can you reopen the issue?

[jkimbo]

Yep

[olejorgenb]

This is an issue AFAIK for all special float values (not really numpy specific) nan, inf, -inf. It's really a pain that json lack a standard way of serializing floats.. :/