- What is JSON?
- Converting JSON string to Python object
- Converting JSON file to Python object
- Converting Python object to JSON string
- Writing Python object to a JSON file
- Converting custom Python objects to JSON objects
- Creating Python class objects from JSON objects
JSON is a common standard used by websites and APIs and even natively supported by modern databases such as PostgreSQL. In this article, we’ll present a tutorial on how to handle JSON data with Python
For a detailed explanation, see our blog post.
JSON, or JavaScript Object Notation, is a format that uses text to store data objects:
{"name": "United States", "population": 331002651, "capital": "Washington D.C.", "languages": [ "English", "Spanish" ] }Let’s start with a simple example:
# JSON stringcountry='{"name": "United States", "population": 331002651}'print(type(country))The output of this snippet will confirm that this is indeed a string:
<class'str'>We can call the json.loads() method and provide this string as a parameter.
importjsoncountry='{"name": "United States", "population": 331002651}'country_dict=json.loads(country) print(type(country)) print(type(country_dict))The output of this snippet will confirm that the JSON data, which was a string, is now a Python dictionary.
<class'str'><class'dict'>This dictionary can be accessed as usual:
print(country_dict['name']) # OUTPUT: United StatesIt is important to note here that the json.loads() method will not always return a dictionary. The data type that is returned will depend on the input string. For example, this JSON string will return a list, not a dictionary.
countries = '["United States", "Canada"]' counties_list= json.loads(countries) print(type(counties_list)) # OUTPUT: <class 'list'> Similarly, if the JSON string contains true, it will be converted to Python equivalent boolean value, which is True.
import json bool_string = 'true' bool_type = json.loads(bool_string) print(bool_type) # OUTPUT: True The following table shows JSON objects and the Python data types after conversion. For more details, see Python docs.
Save the following JSON data as a new file and name it united_states.json:
{"name": "United States", "population": 331002651, "capital": "Washington D.C.", "languages": [ "English", "Spanish" ] }Enter this Python script in a new file:
importjsonwithopen('united_states.json') asf: data=json.load(f) print(type(data))Running this Python file prints the following:
<class 'dict'> The dictionary keys can be checked as follows:
print(data.keys()) # OUTPUT: dict_keys(['name', 'population', 'capital', 'languages']) Using this information, the value of name can be printed as follows:
data['name'] # OUTPUT: United StatesSave this code in a new file as a Python script:
importjsonlanguages= ["English","French"] country={"name": "Canada", "population": 37742154, "languages": languages, "president": None, } country_string=json.dumps(country) print(country_string)When this file is run with Python, the following output is printed:
{"name": "Canada", "population": 37742154, "languages": ["English", "French"], "president": null}Lists can be converted to JSON as well. Here is the Python script and its output:
importjsonlanguages= ["English", "French"] languages_string=json.dumps(languages) print(languages_string) # OUTPUT: ["English", "French"]It’s not just limited to a dictionary and a list. string, int, float, bool and even None value can be converted to JSON.
The method used to write a JSON file is dump():
importjson# Tuple is encoded to JSON array.languages= ("English", "French") # Dictionary is encoded to JSON object.country={"name": "Canada", "population": 37742154, "languages": languages, "president": None, } withopen('countries_exported.json', 'w') asf: json.dump(country, f)To make it more readable, we can pass one more parameter to the dump() function as follows:
json.dump(country, f, indent=4)This time when you run the code, it will be nicely formatted with indentation of 4 spaces:
{"languages": [ "English", "French" ], "president": null, "name": "Canada", "population": 37742154 }Save the following code as a Python script and run it:
importjsonclassCountry: def__init__(self, name, population, languages): self.name=nameself.population=populationself.languages=languagescanada=Country("Canada", 37742154, ["English", "French"]) print(json.dumps(canada)) # OUTPUT: TypeError: Object of type Country is not JSON serializableTo convert the objects to JSON, we need to write a new class that extends JSONEncoder:
importjsonclassCountryEncoder(json.JSONEncoder): defdefault(self, o): ifisinstance(o, Country): # JSON object would be a dictionary.return{"name" : o.name, "population": o.population, "languages": o.languages } else: # Base class will raise the TypeError.returnsuper().default(o)This class can now be supplied to the json.dump() as well as json.dumps() methods.
print(json.dumps(canada, cls=CountryEncoder)) # OUTPUT:{“name": "Canada", "population": 37742154, "languages": ["English", "French"]}Using a custom encoder, we were able to write code like this:
# Create an object of class Countrycanada=Country("Canada", 37742154, ["English", "French"]) # Use json.dump() to create a JSON file in writing modewithopen('canada.json','w') asf: json.dump(canada,f, cls=CountryEncoder)If we try to parse this JSON file using the json.load() method, we will get a dictionary:
withopen('canada.json','r') asf: country_object=json.load(f) # OUTPUT: <type ‘dict'>To get an instance of the Country class instead of a dictionary, we need to create a custom decoder:
importjsonclassCountryDecoder(json.JSONDecoder): def__init__(self, object_hook=None, *args, **kwargs): super().__init__(object_hook=self.object_hook, *args, **kwargs) defobject_hook(self, o): decoded_country=Country( o.get('name'), o.get('population'), o.get('languages'), ) returndecoded_countryFinally, we can call the json.load() method and set the cls parameter to CountryDecoder class.
withopen('canada.json','r') asf: country_object=json.load(f, cls=CountryDecoder) print(type(country_object)) # OUTPUT: <class ‘Country'>If you wish to find out more about Reading & Parsing JSON Data With Python, see our blog post.