How to convert nested numpy arrays in a dictionary to JSON?

I’m parsing nested dictionaries, which have different degrees of how nested they are (dictionaries within dictionaries within dictionaries, etc.) I do not know beforehand to what degree the dictionaries are nested.

The problem is, certain dictionary values are numpy.ndarrays. When I try to write the dictionary my_dictionary to JSON with

with open(my_dictionary, 'w') as f:
    json.dump(my_dictionary, f, indent=4) 

I will get the following error:

TypeError: Object of type ndarray is not JSON serializable

Naturally, one way to overcome this would be to simply convert all numpy.ndarray values into a list with .tolist().

However, given I do not know how nested the dictionaries are, how could I algorithmically check all values of any nested dictionary and convert ndarray to list?

Or is there another way to overcome this error?

Answer

You can make custom json.JSONEncoder. For example:

import json


class MyEncoder(json.JSONEncoder):
    def default(self, obj):
        if isinstance(obj, np.ndarray):
            return obj.tolist()
        return json.JSONEncoder.default(self, obj)


data = {"data": [{"obj": np.array([1, 2, 3])}]}

print(json.dumps(data, cls=MyEncoder, indent=4))

Prints:

{
    "data": [
        {
            "obj": [
                1,
                2,
                3
            ]
        }
    ]
}