Deserialize a json string to an object in python

Question

I have the following string

{\"action\":\"print\",\"method\":\"onData\",\"data\":\"Madan Mohan\"}

I Want to deserialize to a object of cla

Answer 1

If you are embracing the type hints in Python 3.6, you can do it like this:

def from_json(data, cls):
    annotations: dict = cls.__annotations__ if hasattr(cls, '__annotations__') else None
    if issubclass(cls, List):
        list_type = cls.__args__[0]
        instance: list = list()
        for value in data:
            instance.append(from_json(value, list_type))
        return instance
    elif issubclass(cls, Dict):
            key_type = cls.__args__[0]
            val_type = cls.__args__[1]
            instance: dict = dict()
            for key, value in data.items():
                instance.update(from_json(key, key_type), from_json(value, val_type))
            return instance
    else:
        instance : cls = cls()
        for name, value in data.items():
            field_type = annotations.get(name)
            if inspect.isclass(field_type) and isinstance(value, (dict, tuple, list, set, frozenset)):
                setattr(instance, name, from_json(value, field_type))
            else:
                setattr(instance, name, value)
        return instance

Which then allows you do instantiate typed objects like this:

class Bar:
    value : int

class Foo:
    x : int
    bar : List[Bar]


obj : Foo = from_json(json.loads('{"x": 123, "bar":[{"value": 3}, {"value": 2}, {"value": 1}]}'), Foo)
print(obj.x)
print(obj.bar[2].value)

This syntax requires Python 3.6 though and does not cover all cases - for example, support for typing.Any... But at least it does not pollute the classes that need to be deserialized with extra init/tojson methods.

Answer 2

Another way is to simply pass the json string as a dict to the constructor of your object. For example your object is:

class Payload(object):
    def __init__(self, action, method, data, *args, **kwargs):
        self.action = action
        self.method = method
        self.data = data

And the following two lines of python code will construct it:

j = json.loads(yourJsonString)
payload = Payload(**j)

Basically, we first create a generic json object from the json string. Then, we pass the generic json object as a dict to the constructor of the Payload class. The constructor of Payload class interprets the dict as keyword arguments and sets all the appropriate fields.

Answer 3

pydantic is an increasingly popular library for python 3.6+ projects. It mainly does data validation and settings management using type hints.

A basic example using different types:

from pydantic import BaseModel

class ClassicBar(BaseModel):
    count_drinks: int
    is_open: bool
 
data = {'count_drinks': '226', 'is_open': 'False'}
cb = ClassicBar(**data)
>>> cb
ClassicBar(count_drinks=226, is_open=False)

What I love about the lib is that you get a lot of goodies for free, like

>>> cb.json()
'{"count_drinks": 226, "is_open": false}'
>>> cb.dict()
{'count_drinks': 226, 'is_open': False}

Answer 4

If you want to save lines of code and leave the most flexible solution, we can deserialize the json string to a dynamic object:

p = lambda:None
p.__dict__ = json.loads('{"action": "print", "method": "onData", "data": "Madan Mohan"}')

>>>> p.action
output: u'print'

>>>> p.method
output: u'onData'

Answer 5

I thought I lose all my hairs for solving this 'challenge'. I faced following problems:

1. How to deserialize nested objects, lists etc.
2. I like constructors with specified fields
3. I don't like dynamic fields
4. I don't like hacky solutions

I found a library called jsonpickle which is has proven to be really useful.

Installation:

pip install jsonpickle

Here is a code example with writing nested objects to file:

import jsonpickle


class SubObject:
    def __init__(self, sub_name, sub_age):
        self.sub_name = sub_name
        self.sub_age = sub_age


class TestClass:

    def __init__(self, name, age, sub_object):
        self.name = name
        self.age = age
        self.sub_object = sub_object


john_junior = SubObject("John jr.", 2)

john = TestClass("John", 21, john_junior)

file_name = 'JohnWithSon' + '.json'

john_string = jsonpickle.encode(john)

with open(file_name, 'w') as fp:
    fp.write(john_string)

john_from_file = open(file_name).read()

test_class_2 = jsonpickle.decode(john_from_file)

print(test_class_2.name)
print(test_class_2.age)
print(test_class_2.sub_object.sub_name)

Output:

John
21
John jr.

Website: http://jsonpickle.github.io/

Hope it will save your time (and hairs).

Answer 6

To elaborate on Sami's answer:

From the docs:

class Payload(object):
    def __init__(self, action, method, data):
        self.action = action
        self.method = method
        self.data = data

import json

def as_payload(dct):
    return Payload(dct['action'], dct['method'], dct['data'])

payload = json.loads(message, object_hook = as_payload)

My objection to the

.__dict__ 

solution is that while it does the job and is concise, the Payload class becomes totally generic - it doesn't document its fields.

For example, if the Payload message had an unexpected format, instead of throwing a key not found error when the Payload was created, no error would be generated until the payload was used.