json-np 0.0.6

Json serialization extended to dates, numpy arrays, and more

json_np

Json for numpy, or if you prefer, json no problems!

Authors: Luca de Alfaro ([e-mail]) and Massimo Di Pierro ([e-mail])

This is a version of Json that can handle also:

• Dates
• Sets
• numpy array
• bytes

There are two ways to use it.

Basic Usage

import json_np
import numpy as np

a = np.array([3, 4, 5])
s = json_np.dumps(a)
aa = json_np.loads(s)

Here, we have applied json_np.dumps to a numpy array, but we could equally well have used a datetime, a dictionary, a list, etc (nesting of such types is allowed).

Class-Based Usage

You can also dump any object. All object attributes, except those whose name begins with underscore (_), will be serialized. Optionally, if serializable_only is True, only objects that are instances of Serializable, or that are dates, numpy arrays, will be serialized.

The deserialization happens in the same class, if:

• the class can be loaded
• the class initializer has no required arguments.

If any of these two conditions is not met, the deserialization happens using the Serializable class.

from json_np import Serializable

class C(object):

    def __init__(self, a=8):
        super().__init__()
        self.a = a # Serialized
        self._b = 32 # Not serialized

c = C("hello")
s = json_np.dumps(c)
cc = json_np.loads(s)

Obviously, you should ensure that C contains compatible fields when it is deserialized, in case the code changed in the meantime.

History

This package was originally developed by Luca and Massimo at Camio, for Python2. The work was open sourced. The package was later ported to Python 3. The authors thank Camio for allowing the open-sourcing of the code.