Serialize Object to JSON

Question

How do I convert an object to JSON?

I even found some codes on the internet that even worked, but my object has some fields that are other objects and so the conversion did not reach the last level.

Example:

class Pessoa(object):
    codigo = 0
    nome = ''
    endereco = Endereco()

class Endereco(object):
    codigo = 0
    logradouro = ''
    bairro = ''
    cidade = Cidade()

class Cidade(object):
    codigo = 0
    nome = ''
    uf = ''

These are my objects, but doing the conversion I found, the result JSON comes only with the values contained in Person in Addressee, then stay City outside the JSON:

{
    'codigo': 1,
    'nome': 'Oliveira',
    'endereco': {
        'codigo': 5,
        'logradouro': 'Rua A',
        'bairro': 'Campos'            
    }
}

I mean, I believe the code didn’t go through all the levels.

I’m using the pyMongo and I saw that it has a library bson that has some conversions, but I did not know how to use it or else it does not do what I need.

Answer 1

First, you need to define your fields within the __init__, and not so in the class, otherwise they will be shared by all instances of the class (i.e. they will be "static") instead of having one for each object:

class Pessoa(object):
    def __init__(self):
        self.codigo = 0
        self.nome = ''
        self.endereco = Endereco()

class Endereco(object):
    def __init__(self):
        self.codigo = 0
        self.logradouro = ''
        self.bairro = ''
        self.cidade = Cidade()

class Cidade(object):
    def __init__(self):
        self.codigo = 0
        self.nome = ''
        self.uf = ''

(can put parameters - optional or mandatory - in the constructor if you want, instead of starting everything with a default value; just be careful with the default values that are also objects, putting it in the method signature will bring unwanted effects)

Many times you don’t accurate class - use literals for dict, or raise them by hand (such as suggested by Christian Felipe) might be enough. However, classes can be useful if you want to give them methods, so I don’t say don’t use, yes if necessary/convenient.

Once this is done, the object/json conversion libraries you experienced should give the correct result. But if you want a simple method to do by hand, you can take advantage of the fact that the field __dict__ of an object transforms its properties into fields of a dict:

import json

def para_dict(obj):
    # Se for um objeto, transforma num dict
    if hasattr(obj, '__dict__'):
        obj = obj.__dict__

    # Se for um dict, lê chaves e valores; converte valores
    if isinstance(obj, dict):
        return { k:para_dict(v) for k,v in obj.items() }
    # Se for uma lista ou tupla, lê elementos; também converte
    elif isinstance(obj, list) or isinstance(obj, tuple):
        return [para_dict(e) for e in obj]
    # Se for qualquer outra coisa, usa sem conversão
    else: 
        return obj

p = Pessoa()
s = json.dumps(para_dict(p))

This should work for the most common cases, and the use of json.dumps ensures that at least nothing invalid will be echoed to the output.

Answer 2

So man, in Python there is the type of data Dictionary which is kind of the Map of other languages, is a structure of (key, value), that is, almost similar to the problem you are needing, because the format Json follows this pattern of (key, value).
Below is an example of a use of this structure that already comes in the language and does not need other packages and modules, but you must adapt to its use there, obviously:

Pessoa = {}
Endereco = {}
Cidade = {}

Endereco['Logradouro'] = "Rua A"
Endereco['Numero'] = 80
Endereco['Bairro'] = "Centro"

Cidade['Nome'] = "Belo Horizonte"
Cidade['UF'] = "Minas Gerais"
Cidade['Codigo'] = 0

Pessoa['Nome'] = "Meu nome é esse"
Pessoa['Codigo'] = 1
Pessoa['Endereco'] = Endereco
Pessoa['Cidade'] = Cidade

print(repr(Pessoa))

Upshot

{
    "Nome": "Meu nome é esse",
    "Codigo": 1,
    "Endereco": {
        "Numero": 80,
        "Logradouro": "Rua A",
        "Bairro": "Centro"
     },
    "Cidade": {
        "Nome": "Belo Horizonte",
        "Codigo": 0,
        "UF": "Minas Gerais"
     }
}

Then you can make a class that assembles these dictionaries separately, using the methods of access to their objects, and, with the method repr(obj) that you already return the json that you need, is solved your problem.