Developing RESTful APIs with Python and Flask

TL;DR: Throughout this article, we will use Flask and Python to develop a RESTful API. We will create an endpoint that returns static data (dictionaries). Afterward, we will create a class with two specializations and a few endpoints to insert and retrieve instances of these classes. Finally, we will look at how to run the API on a Docker container. The final code developed throughout this article is available in this GitHub repository. I hope you enjoy it!

“Flask allows Python developers to create lightweight RESTful APIs.”

Tweet This

Summary

This article is divided into the following sections:

1. Why Python?
2. Why Flask?
3. Bootstrapping a Flask Application
4. Creating a RESTful Endpoint with Flask
5. Mapping Models with Python Classes
6. Serializing and Deserializing Objects with Marshmallow
7. Dockerizing Flask Applications
8. Securing Python APIs with Auth0
9. Next Steps

Why Python?

Nowadays, choosing Python to develop applications is becoming a very popular choice. As StackOverflow recently analyzed, Python is one of the fastest-growing programming languages, having surpassed even Java in the number of questions asked on the platform. On GitHub, the language also shows signs of mass adoption, occupying the second position among the top programming languages in 2021.

The huge community forming around Python is improving every aspect of the language. More and more open source libraries are being released to address many different subjects, like Artificial Intelligence, Machine Learning, and web development. Besides the tremendous support provided by the overall community, the Python Software Foundation also provides excellent documentation, where new adopters can learn its essence fast.

Why Flask?

When it comes to web development on Python, there are three predominant frameworks: Django, Flask, and a relatively new player FastAPI. Django is older, more mature, and a little bit more popular. On GitHub, this framework has around 66k stars, 2.2k contributors, ~ 350 releases, and more than 25k forks.

FastAPI is growing at high speed, with 48k stars on Github, 370 contributors, and more than 3.9k forks. This elegant framework built for high-performance and fast-to-code APIs is not one to miss.

Flask, although less popular, is not far behind. On GitHub, Flask has almost 60k stars, ~650 contributors, ~23 releases, and nearly 15k forks.

Even though Django is older and has a slightly more extensive community, Flask has its strengths. From the ground up, Flask was built with scalability and simplicity. Flask applications are known for being lightweight, mainly compared to their Django counterparts. Flask developers call it a microframework, where micro (as explained here) means that the goal is to keep the core simple but extensible. Flask won't make many decisions for us, such as what database to use or what template engine to choose. Lastly, Flask has extensive documentation that addresses everything developers need to start. FastAPI follows a similar "micro" approach to Flask, though it provides more tools like automatic Swagger UI and is an excellent choice for APIs. However, as it is a newer framework, many more resources and libraries are compatible with frameworks like Django and Flask but not with FastAPI.

Being lightweight, easy to adopt, well-documented, and popular, Flask is a good option for developing RESTful APIs.

Bootstrapping a Flask Application

First and foremost, we will need to install some dependencies on our development machine. We will need to install Python 3, Pip (Python Package Index), and Flask.

Installing Python 3

If we are using some recent version of a popular Linux distribution (like Ubuntu) or macOS, we might already have Python 3 installed on our computer. If we are running Windows, we will probably need to install Python 3, as this operating system does not ship with any version.

After installing Python 3 on our machine, we can check that we have everything set up as expected by running the following command:

python --version
# Python 3.8.9

Note that the command above might produce a different output when we have a different Python version. What is important is that you are running at least

Python 3.7

python3 --version

python3

python

Installing Pip

Pip is the recommended tool for installing Python packages. While the official installation page states that

pip

2.7.9

3.4

apt

pip

pip

# we might need to change pip by pip3
pip --version
# pip 9.0.1 ... (python 3.X)

If the command above produces an output similar to

pip 9.0.1 ... (python 3.X)

pip 9.0.1 ... (python 2.X)

pip

pip3

Installing Flask

We already know what Flask is and its capabilities. Therefore, let's focus on installing it on our machine and testing to see if we can get a basic Flask application running. The first step is to use

pip

# we might need to replace pip with pip3
pip install Flask

After installing the package, we will create a file called

hello.py

# hello.py

from flask import Flask

app = Flask(__name__)

@app.route("/")
def hello_world():
    return "Hello, World!"

These 5 lines of code are everything we need to handle HTTP requests and return a "Hello, World!" message. To run it, we execute the following command:

flask --app hello run

 * Serving Flask app 'hello'
 * Debug mode: off
WARNING: This is a development server. Do not use it in a production deployment. Use a production WSGI server instead.
 * Running on http://127.0.0.1:5000
Press CTRL+C to quit

On Ubuntu, we might need to edit the

$PATH

variable to be able to run flask directly. To do that, let's

touch ~/.bash_aliases

and then

echo "export PATH=$PATH:~/.local/bin" >> ~/.bash_aliases

.

After executing these commands, we can reach our application by opening a browser and navigating to

http://127.0.0.1:5000/

curl http://127.0.0.1:5000/

Virtual environments (virtualenv)

Although PyPA—the Python Packaging Authority group—recommends

pip

pip

requirements.txt

• pip

  installs packages globally, making it hard to manage multiple versions of the same package on the same machine.

• requirements.txt

  need all dependencies and sub-dependencies listed explicitly, a manual process that is tedious and error-prone.

To solve these issues, we are going to use Pipenv. Pipenv is a dependency manager that isolates projects in private environments, allowing packages to be installed per project. If you're familiar with NPM or Ruby's bundler, it's similar in spirit to those tools.

pip install pipenv

Now, to start creating a serious Flask application, let's create a new directory that will hold our source code. In this article, we will create Cashman, a small RESTful API that allows users to manage incomes and expenses. Therefore, we will create a directory called

cashman-flask-project

pipenv

# create our project directory and move to it
mkdir cashman-flask-project && cd cashman-flask-project

# use pipenv to create a Python 3 (--three) virtualenv for our project
pipenv --three

# install flask a dependency on our project
pipenv install flask

The second command creates our virtual environment, where all our dependencies get installed, and the third will add Flask as our first dependency. If we check our project's directory, we will see two new files:

1. Pipfile

   contains details about our project, such as the Python version and the packages needed.

2. Pipenv.lock

   contains precisely what version of each package our project depends on and its transitive dependencies.

Python modules

Like other mainstream programming languages, Python also has the concept of modules to enable developers to organize source code according to subjects/functionalities. Similar to Java packages and C# namespaces, modules in Python are files organized in directories that other Python scripts can import. To create a module on a Python application, we need to create a folder and add an empty file called

__init__.py

Let's create our first module on our application, the main module, with all our RESTful endpoints. Inside the application's directory, let's create another one with the same name,

cashman

cashman-flask-project

# create source code's root
mkdir cashman && cd cashman

# create an empty __init__.py file
touch __init__.py

Inside the main module, let's create a script called

index.py

from flask import Flask
app = Flask(__name__)


@app.route("/")
def hello_world():
    return "Hello, World!"

As in the previous example, our application returns a "Hello, world!" message. We will start improving it in a second, but first, let's create an executable file called

bootstrap.sh

# move to the root directory
cd ..

# create the file
touch bootstrap.sh

# make it executable
chmod +x bootstrap.sh

The goal of this file is to facilitate the start-up of our application. Its source code will be the following:

#!/bin/sh
export FLASK_APP=./cashman/index.py
pipenv run flask --debug run -h 0.0.0.0

The first command defines the main script to be executed by Flask. The second command runs our Flask application in the context of the virtual environment listening to all interfaces on the computer (

-h 0.0.0.0

Note: we are setting flask to run in debug mode to enhance our development experience and activate the hot reload feature, so we don't have to restart the server each time we change the code. If you run Flask in production, we recommend updating these settings for production.

To check that this script is working correctly, we run

./bootstrap.sh

 * Serving Flask app './cashman/index.py'
 * Debug mode: on
WARNING: This is a development server. Do not use it in a production deployment. Use a production WSGI server instead.
 * Running on all addresses (0.0.0.0)
 * Running on http://127.0.0.1:5000
 * Running on http://192.168.1.207:5000
Press CTRL+C to quit

Creating a RESTful Endpoint with Flask

Now that our application is structured, we can start coding some relevant endpoints. As mentioned before, the goal of our application is to help users to manage incomes and expenses. We will begin by defining two endpoints to handle incomes. Let's replace the contents of the

./cashman/index.py

from flask import Flask, jsonify, request

app = Flask(__name__)

incomes = [
    { 'description': 'salary', 'amount': 5000 }
]


@app.route('/incomes')
def get_incomes():
    return jsonify(incomes)


@app.route('/incomes', methods=['POST'])
def add_income():
    incomes.append(request.get_json())
    return '', 204

Since improving our application, we have removed the endpoint that returned "Hello, world!" to users. In its place, we defined an endpoint to handle HTTP

GET

POST

@app.route

/incomes

To facilitate the process, we currently manipulate incomes as dictionaries. However, we will soon create classes to represent incomes and expenses.

To interact with both endpoints that we have created, we can start our application and issue some HTTP requests:

# start the cashman application
./bootstrap.sh &

# get incomes
curl http://localhost:5000/incomes

# add new income
curl -X POST -H "Content-Type: application/json" -d '{
  "description": "lottery",
  "amount": 1000.0
}' http://localhost:5000/incomes

# check if lottery was added
curl localhost:5000/incomes

Mapping Models with Python Classes

Using dictionaries in a simple use case like the one above is enough. However, for more complex applications that deal with different entities and have multiple business rules and validations, we might need to encapsulate our data into Python classes.

We will refactor our application to learn the process of mapping entities (like incomes) as classes. The first thing that we will do is create a submodule to hold all our entities. Let's create a

model

cashman

__init__.py

# create model directory inside the cashman module
mkdir -p cashman/model

# initialize it as a module
touch cashman/model/__init__.py

Mapping a Python superclass

We will create three classes in this new module/directory:

Transaction

Income

Expense

Transaction

transaction.py

model

import datetime as dt

from marshmallow import Schema, fields


class Transaction(object):
    def __init__(self, description, amount, type):
        self.description = description
        self.amount = amount
        self.created_at = dt.datetime.now()
        self.type = type

    def __repr__(self):
        return '<Transaction(name={self.description!r})>'.format(self=self)


class TransactionSchema(Schema):
    description = fields.Str()
    amount = fields.Number()
    created_at = fields.Date()
    type = fields.Str()

Besides the

Transaction

TransactionSchema

Transaction

Schema

# installing marshmallow as a project dependency
pipenv install marshmallow

Marshmallow is a popular Python package for converting complex datatypes, such as objects, to and from built-in Python datatypes. We can use this package to validate, serialize, and deserialize data. We won't dive into validation in this article, as it will be the subject of another one. Though, as mentioned, we will use

marshmallow

Mapping Income and Expense as Python Classes

To keep things more organized and meaningful, we won't expose the

Transaction

Income

Expense

income.py

model

from marshmallow import post_load

from .transaction import Transaction, TransactionSchema
from .transaction_type import TransactionType


class Income(Transaction):
    def __init__(self, description, amount):
        super(Income, self).__init__(description, amount, TransactionType.INCOME)

    def __repr__(self):
        return '<Income(name={self.description!r})>'.format(self=self)


class IncomeSchema(TransactionSchema):
    @post_load
    def make_income(self, data, **kwargs):
        return Income(**data)

The only value that this class adds for our application is that it hardcodes the type of transaction. This type is a Python enumerator, which we still have to create, that will help us filter transactions in the future. Let's create another file, called

transaction_type.py

model

from enum import Enum


class TransactionType(Enum):
    INCOME = "INCOME"
    EXPENSE = "EXPENSE"

The code of the enumerator is quite simple. It just defines a class called

TransactionType

Enum

INCOME

EXPENSE

Lastly, let's create the class that represents expenses. To do that, let's add a new file called

expense.py

model

from marshmallow import post_load

from .transaction import Transaction, TransactionSchema
from .transaction_type import TransactionType


class Expense(Transaction):
    def __init__(self, description, amount):
        super(Expense, self).__init__(description, -abs(amount), TransactionType.EXPENSE)

    def __repr__(self):
        return '<Expense(name={self.description!r})>'.format(self=self)


class ExpenseSchema(TransactionSchema):
    @post_load
    def make_expense(self, data, **kwargs):
        return Expense(**data)

Similar to

Income

EXPENSE

amount

Serializing and Deserializing Objects with Marshmallow

With the

Transaction

./cashman/index.py

from flask import Flask, jsonify, request

from cashman.model.expense import Expense, ExpenseSchema
from cashman.model.income import Income, IncomeSchema
from cashman.model.transaction_type import TransactionType

app = Flask(__name__)

transactions = [
    Income('Salary', 5000),
    Income('Dividends', 200),
    Expense('pizza', 50),
    Expense('Rock Concert', 100)
]


@app.route('/incomes')
def get_incomes():
    schema = IncomeSchema(many=True)
    incomes = schema.dump(
        filter(lambda t: t.type == TransactionType.INCOME, transactions)
    )
    return jsonify(incomes)


@app.route('/incomes', methods=['POST'])
def add_income():
    income = IncomeSchema().load(request.get_json())
    transactions.append(income)
    return "", 204


@app.route('/expenses')
def get_expenses():
    schema = ExpenseSchema(many=True)
    expenses = schema.dump(
        filter(lambda t: t.type == TransactionType.EXPENSE, transactions)
    )
    return jsonify(expenses)


@app.route('/expenses', methods=['POST'])
def add_expense():
    expense = ExpenseSchema().load(request.get_json())
    transactions.append(expense)
    return "", 204


if __name__ == "__main__":
    app.run()

The new version that we just implemented starts by redefining the

incomes

Expenses

Incomes

transactions

IncomeSchema

filter

transactions

The endpoint responsible for accepting new incomes was also refactored. The change on this endpoint was the addition of

IncomeSchema

Income

transactions

Transaction

Income

The other two endpoints responsible for dealing with expenses,

get_expenses

add_expense

income

• instead of dealing with instances of

  Income

  , we deal with instances of

  Expense

  to accept new expenses,
• and instead of filtering by

  TransactionType.INCOME

  , we filter by

  TransactionType.EXPENSE

  to send expenses back to the user.

This finishes the implementation of our API. If we run our Flask application now, we will be able to interact with the endpoints, as shown here:

# start the application
./bootstrap.sh

# get expenses
curl http://localhost:5000/expenses

# add a new expense
curl -X POST -H "Content-Type: application/json" -d '{
    "amount": 20,
    "description": "lottery ticket"
}' http://localhost:5000/expenses

# get incomes
curl http://localhost:5000/incomes

# add a new income
curl -X POST -H "Content-Type: application/json" -d '{
    "amount": 300.0,
    "description": "loan payment"
}' http://localhost:5000/incomes

Dockerizing Flask Applications

As we are planning to eventually release our API in the cloud, we are going to create a

Dockerfile

Dockerfile

Let's create the

Dockerfile

# Using lightweight alpine image
FROM python:3.8-alpine

# Installing packages
RUN apk update
RUN pip install --no-cache-dir pipenv

# Defining working directory and adding source code
WORKDIR /usr/src/app
COPY Pipfile Pipfile.lock bootstrap.sh ./
COPY cashman ./cashman

# Install API dependencies
RUN pipenv install --system --deploy

# Start app
EXPOSE 5000
ENTRYPOINT ["/usr/src/app/bootstrap.sh"]

The first item in the recipe defines that we will create our Docker container based on the default Python 3 Docker image. After that, we update APK and install

pipenv

pipenv

pipenv

5000

bootstrap.sh

Note: For our

Dockerfile

, we use Python version 3.8, however, depending on your system configuration,

pipenv

may have set a different version for Python in the file

Pipfile

. Please make sure that the Python version in both

Dockerfile

and

Pipfile

are aligned, or the docker container won't be able to start the server.

To create and run a Docker container based on the

Dockerfile

# build the image
docker build -t cashman .

# run a new docker container named cashman
docker run --name cashman \
    -d -p 5000:5000 \
    cashman

# fetch incomes from the dockerized instance
curl http://localhost:5000/incomes/

The

Dockerfile

Dockerfile

requires_auth

# Format error response and append status code

def get_token_auth_header():
    """Obtains the access token from the Authorization Header
    """
    auth = request.headers.get("Authorization", None)
    if not auth:
        raise AuthError({"code": "authorization_header_missing",
                        "description":
                            "Authorization header is expected"}, 401)

    parts = auth.split()

    if parts[0].lower() != "bearer":
        raise AuthError({"code": "invalid_header",
                        "description":
                            "Authorization header must start with"
                            " Bearer"}, 401)
    elif len(parts) == 1:
        raise AuthError({"code": "invalid_header",
                        "description": "Token not found"}, 401)
    elif len(parts) > 2:
        raise AuthError({"code": "invalid_header",
                        "description":
                            "Authorization header must be"
                            " Bearer token"}, 401)

    token = parts[1]
    return token

def requires_auth(f):
    """Determines if the access token is valid
    """
    @wraps(f)
    def decorated(*args, **kwargs):
        token = get_token_auth_header()
        jsonurl = urlopen("https://"+AUTH0_DOMAIN+"/.well-known/jwks.json")
        jwks = json.loads(jsonurl.read())
        unverified_header = jwt.get_unverified_header(token)
        rsa_key = {}
        for key in jwks["keys"]:
            if key["kid"] == unverified_header["kid"]:
                rsa_key = {
                    "kty": key["kty"],
                    "kid": key["kid"],
                    "use": key["use"],
                    "n": key["n"],
                    "e": key["e"]
                }
        if rsa_key:
            try:
                payload = jwt.decode(
                    token,
                    rsa_key,
                    algorithms=ALGORITHMS,
                    audience=API_AUDIENCE,
                    issuer="https://"+AUTH0_DOMAIN+"/"
                )
            except jwt.ExpiredSignatureError:
                raise AuthError({"code": "token_expired",
                                "description": "token is expired"}, 401)
            except jwt.JWTClaimsError:
                raise AuthError({"code": "invalid_claims",
                                "description":
                                    "incorrect claims,"
                                    "please check the audience and issuer"}, 401)
            except Exception:
                raise AuthError({"code": "invalid_header",
                                "description":
                                    "Unable to parse authentication"
                                    " token."}, 400)

            _app_ctx_stack.top.current_user = payload
            return f(*args, **kwargs)
        raise AuthError({"code": "invalid_header",
                        "description": "Unable to find appropriate key"}, 400)
    return decorated

# Controllers API

# This doesn't need authentication
@app.route("/ping")
@cross_origin(headers=['Content-Type', 'Authorization'])
def ping():
    return "All good. You don't need to be authenticated to call this"

# This does need authentication
@app.route("/secured/ping")
@cross_origin(headers=['Content-Type', 'Authorization'])
@requires_auth
def secured_ping():
    return "All good. You only get this message if you're authenticated"

Next Steps

In this article, we learned about the basic components needed to develop a well-structured Flask application. We looked at how to use

pipenv

Although well structured, our API is not that useful yet. Among the things that we can improve, we are going to cover the following topics in the following article:

• Database persistence with SQLAlchemy
• Add authorization to a Flask API application
• How to handle JWTs in Python

Stay tuned!