Invalid JSON being generated by save_json methods

[jlopezpena]

The JSON encoder used can output NaN and infinite values (see the most_common_value label in the example below).

This is valid javascript, but not valid JSON, and can create problems down the pipeline when another service tries to deserialize it.

The standard json.dumps method has an allow_nan parameter that can be used to control this behaviour, and it would be good to have access to it (with the default being left as it is right now for compatibility).

Generally speaking, it would be good to have access to all the parameters used by json.dumps, perhaps by adding some catch-all **kwargs to the .json() method in evidently and passing that along to the json.dumps call.

Example:

{
    "feature_type": "num",
    "count": 2492,
    "infinite_count": 0,
    "infinite_percentage": 0.0,
    "missing_count": 190,
    "missing_percentage": 7.08,
    "unique_count": 2482,
    "unique_percentage": 92.54,
    "percentile_25": -0.11,
    "percentile_50": 0.35,
    "percentile_75": 1.41,
    "max": 241887.34,
    "min": -1.0,
    "mean": 138.11,
    "most_common_value": NaN,
    "most_common_value_percentage": 7.08,
    "std": 4949.03,
    "most_common_not_null_value": 1.098466433173,
    "most_common_not_null_value_percentage": 0.07,
    "new_in_current_values_count": null,
    "unused_in_current_values_count": null
}

[emeli-dral]

Hi @jlopezpena, thanks for your suggestion, it is a great point. We were unable to fit in the current release as we focused on major refactoring. But we took it into work and will make the changes in the next release in November.