Python Library: JSON

JSON

JSON is JavaScript Object Notation and it is a data interchange format used widely on the web and elsewhere. The nature of JavaScript is dynamic, and Python works well for parsing and working with JSON. The standard library of Python has a wonderful JSON module that makes working with JSON simple.

JSON to Python

JSON maps nicely to Python objects when parsed. When translating Python to JSON, this process is reversible. However, more complex Python objects like datetime or namedtuples will not translate without custom behavior specified.

JSON	Python
String	String
Number	Int or Float
Object	Dictionary
Array	List
null	None
true	True
false	False

Loads and Dumps

The main operations used with the JSON library are loads and dumps. loads takes in a String of JSON and parses it into the corresponding Python objects. dumps is the reverse, it takes a Python object and creates a String of JSON. You’ll often receive JSON responses from web servers and parse them with loads. Then you can create a Python object and stringify it with dumps to send back.

import json

# I like using single quotes, especially for JSON strings
# because then you don't need to escape double quotes
json_string = '{"key": "value", "array_key": [1, 2, 3], "another_key_for_boolean": false}'

python_object = json.loads(json_string)
print(python_object)
# {'key': 'value', 'array_key': [1, 2, 3], 'another_key_for_boolean': False}

Now we can operate on the Python object with our normal Python syntax. Once we are done, we can create a new JSON string.

python_object['new_key'] = None
python_object['array_key'].append(4)

new_json_string = json.dumps(python_object)

print(new_json_string)
# '{"key": "value", "array_key": [1, 2, 3, 4], "another_key_for_boolean": false, "new_key": null}'

Another other nice features of dumps is the ability to pretty print with the indent parameter. Just specify a number for the indentation, and the JSON string will be nicely formatted for human readable output.

import json

python_object = {'parent': {'child_1': [1, 2, 3], 'child_2': True}}

json.dumps(python_object, indent=4)

{
    "parent": {
        "child_1": [
            1,
            2,
            3
        ],
        "child_2": true
    }
}

Overriding default dumps behavior

Some objects are not directly translatable into JSON. datetime objects are a great example and will come up often.

import datetime

now = datetime.datetime.now()
# datetime.datetime(2020, 8, 3, 20, 35, 38, 299215)

json.dumps(now)
# TypeError: Object of type datetime is not JSON serializable

The solution to creating JSON strings with datetime and other complex objects is to subclass the JSONEncoder class and implement a default method that returns a Python object that can be encoded. For a datetime object, turn it into a dictionary with a key of ‘datetime_now’ and a string representation of the datetime object, which is JSON serializable.

import datetime
import json

now = datetime.datetime.now()

class DateTimeEncoder(json.JSONEncoder):
    def default(self, o):
        if type(o) == datetime.datetime:
            return {'datetime_now': str(o)}
        return json.JSONEncoder.default(self, o)

date_time_encoder = DateTimeEncoder()

now_json_string = date_time_encoder.encode(now)

print(now_json_string)
# '{"datetime_now": "2020-08-03 20:35:38.299215"}'