LGTM. Do you want to add "NaN" in integration/data/simple.json?

The background for this is that an existing Spark unit test for Arrow fails with the current master branch due to no longer being able to parse these values. The configuration fixes this and allows for the following JSON:

    {
      "name" : "float",
      "count" : 10,
      "VALIDITY" : [1,1,1,1,1,1,1,1,1,1],
      "DATA" : [NaN,1.0,2.0,3.0,4.0,5.0,6.0,7.0,8.0,9.0]
    }

This is not standard JSON, but is supported by the Jackson library, I don't think it is something to be concerned about though. cc @icexelloss

LGTM. Do you want to add "NaN" in integration/data/simple.json?

Thanks @icexelloss for the quick review! Yeah, I'll try adding this to the integration tests and see if it's handled

It doesn't look like C++ is able to parse the NaN value, but I'm not sure if this is something that we need to be running integration for. @wesm what do you think?

Hm, I can try to fix the C++ reader. Up to you

Given that the issue was only on the Java side and it's not standard JSON, I would say not to bother with the C++ reader and integration tests for now.

LGTM. +1. With or without NaN in integration test.

Thanks @icexelloss , I'll try merging this later today if no more comments