This articles serves as a guide to testing Flask applications with pytest.
We'll first look at why testing is important for creating maintainable software and what you should focus on when testing. Then, we'll detail how to:
- Create and run unit and functional tests with pytest
- Utilize fixtures to initialize the state for test functions
- Check the coverage of the tests using coverage.py
The source code (along with detailed installation instructions) for the Flask app being tested in this article can be found on GitLab at https://gitlab.com/patkennedy79/flask_user_management_example.
By the end of this article, you should be able to:
- Explain what to test in a Flask app
- Describe the differences between pytest and unittest
- Write unit and functional test functions with pytest
- Run tests with pytest
- Create fixtures for initializing the state for test functions
- Determine code coverage of your tests with coverage.py
Why Write Tests?
In general, testing helps ensure that your app will work as expected for your end users.
Software projects with high test coverage are never perfect, but it's a good initial indicator of the quality of the software. Additionally, testable code is generally a sign of a good software architecture, which is why advanced developers take testing into account throughout the entire development lifecycle.
Tests can be considered at three levels:
- Functional (or integration)
Unit tests test the functionality of an individual unit of code isolated from its dependencies. They are the first line of defense against errors and inconsistencies in your codebase. They test from the inside out, from the programmer's point of view.
Functional tests test multiple components of a software product to make sure the components are working together properly. Typically, these tests are focused on functionality that the user will be utilizing. They test from the outside in, from the end user's point of view.
Both unit and Functional testing are fundamental parts of the Test-Driven Development (TDD) process.
Testing improves the maintainability of your code.
Maintainability refers to making bug fixes or enhancements to your code or to another developer needing to update your code at some point in the future.
Testing should be combined with a Continuous Integration (CI) process to ensure that your tests are constantly executing, ideally on each commit to your repository. A solid suite of tests can be critical to catching defects quickly and early in the development process before your end users come across them in production.
What to Test?
What should you test?
Again, unit tests should focus on testing small units of code in isolation.
For example, in a Flask app, you may use unit tests to test:
- Database models
- Utility functions that your view functions call
Functional tests, meanwhile, should focus on how the view functions operate.
- Nominal conditions (GET, POST, etc.) for a view function
- Invalid HTTP methods are handled properly for a view function
- Invalid data is passed to a view function
Focus on testing scenarios that the end user will interact with. The experience that the users of your product have is paramount!
pytest vs. unittest
pytest is a test framework for Python used to write, organize, and run test cases. After setting up your basic test structure, pytest makes it really easy to write tests and provides a lot of flexibility for running the tests. pytest satisfies the key aspects of a good test environment:
- tests are fun to write
- tests can be written quickly by using helper functions (fixtures)
- tests can be executed with a single command
- tests run quickly
pytest is incredible! I highly recommend using it for testing any application or script written in Python.
If you're interested in really learning all the different aspects of pytest, I highly recommend the Python Testing with pytest book by Brian Okken.
It provides the following:
- tools for building unit tests, including a full suite of
assertstatements for performing checks
- structure for developing unit tests and unit test suites
- test runner for executing tests
The main differences between pytest and unittest:
|Installation||Third-party library||Part of the core standard library|
|Test setup and teardown||fixtures||
|Assertion Format||Built-in assert||
Either framework is good for testing a Flask project. However, I prefer pytest since it:
- Requires less boilerplate code so your test suites will be more readable.
- Supports the plain
assertstatement, which is far more readable and easier to remember compared to the
assertSomethingmethods -- like
assertContains-- in unittest.
- Is updated more frequently since it's not part of the Python standard library.
- Simplifies setting up and tearing down test state.
- Uses a functional approach.
- Supports fixtures.
I like to organize all the test cases in a separate "tests" folder at the same level as the application files.
Additionally, I really like differentiating between unit and functional tests by splitting them out as separate sub-folders. This structure gives you the flexibility to easily run just the unit tests (or just the functional tests, for that matter).
Here's an example of the structure of the "tests" directory:
└── tests ├── conftest.py ├── functional │ ├── __init__.py │ ├── test_stocks.py │ └── test_users.py └── unit ├── __init__.py └── test_models.py
And, here's how the "tests" folder fits into a typical Flask project with blueprints:
├── app.py ├── project │ ├── __init__.py │ ├── models.py │ └── ...blueprint folders... ├── requirements.txt ├── tests │ ├── conftest.py │ ├── functional │ │ ├── __init__.py │ │ ├── test_stocks.py │ │ └── test_users.py │ └── unit │ ├── __init__.py │ └── test_app.py └── venv
Unit Test Example
This test doesn't access the underlying database; it only checks the interface class used by SQLAlchemy.
Since this test is a unit test, it should be implemented in tests/unit/test_models.py:
from project.models import User def test_new_user(): """ GIVEN a User model WHEN a new User is created THEN check the email, hashed_password, and role fields are defined correctly """ user = User('[email protected]', 'FlaskIsAwesome') assert user.email == '[email protected]' assert user.hashed_password != 'FlaskIsAwesome' assert user.role == 'user'
Let's take a closer look at this test.
After the import, we start with a description of what the test does:
""" GIVEN a User model WHEN a new User is created THEN check the email, hashed_password, and role fields are defined correctly """
Why include so many comments for a test function?
I've found that tests are one of the most difficult aspects of a project to maintain. Often, the code (including the level of comments) for test suites is nowhere near the level of quality as the code being tested.
A common structure used to describe what each test function does helps with maintainability by making it easier for a someone (another developer, your future self) to quickly understand the purpose of each test.
A common practice is to use the GIVEN-WHEN-THEN structure:
- GIVEN - what are the initial conditions for the test?
- WHEN - what is occurring that needs to be tested?
- THEN - what is the expected response?
Next, we have the actual test:
user = User('[email protected]', 'FlaskIsAwesome') assert user.email == '[email protected]' assert user.hashed_password != 'FlaskIsAwesome' assert user.role == 'user'
After creating a new
user with valid arguments to the constructor, the properties of the
user are checked to make sure it was created properly.
Functional Test Examples
The second test that we're going to write is an functional test for project/recipes/routes.py, which contains the view functions for the
Since this test is a functional test, it should be implemented in tests/functional/test_recipes.py:
from project import create_app def test_home_page(): """ GIVEN a Flask application configured for testing WHEN the '/' page is requested (GET) THEN check that the response is valid """ flask_app = create_app('flask_test.cfg') # Create a test client using the Flask application configured for testing with flask_app.test_client() as test_client: response = test_client.get('/') assert response.status_code == 200 assert b"Welcome to the Flask User Management Example!" in response.data assert b"Need an account?" in response.data assert b"Existing user?" in response.data
This project uses the Application Factory Pattern to create the Flask application. Therefore, the
create_app() function needs to first be imported:
from project import create_app
The test function,
test_home_page(), starts with the GIVEN-WHEN-THEN description of what the test does. Next, a Flask application (
flask_app) is created:
flask_app = create_app('flask_test.cfg')
In order to create the proper environment for testing, Flask provides a test_client helper. This creates a test version of our Flask application, which we used to make a GET call to the '/' URL. We then check that the status code returned is OK (200) and that the response contained the following strings:
- Welcome to the Flask User Management Example!
- Need an account?
- Existing user?
These checks match with what we expect the user to see when we navigate to the '/' URL:
An example of an off-nominal functional test would be to utilize an invalid HTTP method (POST) when accessing the '/' URL:
def test_home_page_post(): """ GIVEN a Flask application configured for testing WHEN the '/' page is is posted to (POST) THEN check that a '405' status code is returned """ flask_app = create_app('flask_test.cfg') # Create a test client using the Flask application configured for testing with flask_app.test_client() as test_client: response = test_client.post('/') assert response.status_code == 405 assert b"Welcome to the Flask User Management Example!" not in response.data
This test checks that a POST request to the '/' URL results in an error code of 405 (Method Not Allowed) being returned.
Take a second to review the two functional tests... do you see some duplicate code between these two test functions? Do you see a lot of code for initializing the state needed by the test functions? We can use fixtures to address these issues.
Fixtures initialize tests to a known state in order to run tests in a predictable and repeatable manner.
The classic approach to writing and executing tests follows the the xUnit type of test framework, where each test runs as follows:
- ...run the test case...
TearDown() methods always run for each unit test within a test suite. This approach results in the same initial state for each test within a test suite, which doesn't provide much flexibility.
Advantages of Fixtures
The test fixture approach provides much greater flexibility than the classic Setup/Teadown approach.
pytest-flask facilitates testing Flask apps by providing a set of common fixtures used for testing Flask apps. This library is not used in this tutorial, as I want to show how to create the fixtures that help support testing Flask apps.
First, fixtures are defined as functions (that should have a descriptive names for their purpose).
Second, multiple fixtures can be run to set the initial state for a test function. In fact, fixtures can even call other fixtures! So, you can compose them together to create the required state.
Finally, fixtures can be run with different scopes:
function- run once per test function (default scope)
class- run once per test class
module- run once per module (e.g., a test file)
session- run once per session
For example, if you have a fixture with module scope, that fixture will run once (and only once) before the test functions in the module run.
Fixtures should be created in tests/conftest.py.
Unit Test Example
from project.models import User @pytest.fixture(scope='module') def new_user(): user = User('[email protected]', 'FlaskIsAwesome') return user
@pytest.fixture decorator specifies that this function is a fixture with
module-level scope. In other words, this fixture will be called one per test module.
new_user, creates an instance of
User using valid arguments to the constructor.
user is then passed to the test function (
We can simplify the
test_new_user() test function from earlier by using the
new_user fixture in tests/unit/test_models.py:
def test_new_user_with_fixture(new_user): """ GIVEN a User model WHEN a new User is created THEN check the email, hashed_password, authenticated, and role fields are defined correctly """ assert new_user.email == '[email protected]' assert new_user.hashed_password != 'FlaskIsAwesome' assert new_user.role == 'user'
By using a fixture, the test function is reduced to the
assert statements that perform the checks against the
Functional Test Examples
To help facilitate testing all the view functions in the Flask project, a fixture can be created in tests/conftest.py:
from project import create_app @pytest.fixture(scope='module') def test_client(): flask_app = create_app('flask_test.cfg') # Create a test client using the Flask application configured for testing with flask_app.test_client() as testing_client: # Establish an application context with flask_app.app_context(): yield testing_client # this is where the testing happens!
This fixture creates the test client using a context manager:
with flask_app.test_client() as testing_client:
Next, the application context is pushed onto the stack for use by the test functions:
with flask_app.app_context(): yield testing_client # this is where the testing happens!
To learn more about the Application context in Flask, refer to the following blog posts:
yield testing_client statement means that execution is being passed to the test functions.
Using the Fixture
We can simplify the functional tests from earlier with the
test_client fixture in tests/functional/test_recipes.py:
def test_home_page_with_fixture(test_client): """ GIVEN a Flask application configured for testing WHEN the '/' page is requested (GET) THEN check that the response is valid """ response = test_client.get('/') assert response.status_code == 200 assert b"Welcome to the Flask User Management Example!" in response.data assert b"Need an account?" in response.data assert b"Existing user?" in response.data def test_home_page_post_with_fixture(test_client): """ GIVEN a Flask application WHEN the '/' page is is posted to (POST) THEN check that a '405' status code is returned """ response = test_client.post('/') assert response.status_code == 405 assert b"Welcome to the Flask User Management Example!" not in response.data
Notice how much duplicate code is eliminated by using the
test_client fixture? By utilizing the
test_client fixture, each test function is simplified down the HTTP call (GET or POST) and the assert that checks the response.
I really find that using fixtures helps to focus the test function on actually doing the testing, as the test initialization is handled in the fixture.
Running the Tests
To run the tests, run
pytest in the top-level folder for the Flask project:
(venv)$ pytest ============================= test session starts ============================== tests/functional/test_recipes.py .... [ 30%] tests/functional/test_users.py ..... [ 69%] tests/unit/test_models.py .... [100%] ============================== 13 passed in 0.46s ==============================
To see more details on the tests that were run:
(venv)$ pytest -v ============================= test session starts ============================== tests/functional/test_recipes.py::test_home_page PASSED [ 7%] tests/functional/test_recipes.py::test_home_page_post PASSED [ 15%] tests/functional/test_recipes.py::test_home_page_with_fixture PASSED [ 23%] tests/functional/test_recipes.py::test_home_page_post_with_fixture PASSED [ 30%] tests/functional/test_users.py::test_login_page PASSED [ 38%] tests/functional/test_users.py::test_valid_login_logout PASSED [ 46%] tests/functional/test_users.py::test_invalid_login PASSED [ 53%] tests/functional/test_users.py::test_valid_registration PASSED [ 61%] tests/functional/test_users.py::test_invalid_registration PASSED [ 69%] tests/unit/test_models.py::test_new_user PASSED [ 76%] tests/unit/test_models.py::test_new_user_with_fixture PASSED [ 84%] tests/unit/test_models.py::test_setting_password PASSED [ 92%] tests/unit/test_models.py::test_user_id PASSED [100%] ============================== 13 passed in 0.62s ==============================
If you only want to run a specific type of test:
Fixtures in Action
To really get a sense of when the
test_client() fixture is run, pytest can provide a call structure of the fixtures and tests with the '--setup-show' argument:
(venv)$ pytest --setup-show tests/functional/test_recipes.py ====================================== test session starts ===================================== tests/functional/test_recipes.py ... SETUP M test_client functional/test_recipes.py::test_home_page_with_fixture (fixtures used: test_client). functional/test_recipes.py::test_home_page_post_with_fixture (fixtures used: test_client). TEARDOWN M test_client ======================================= 4 passed in 0.18s ======================================
test_client fixture has a 'module' scope, so it's executed prior to the two _with_fixture tests in tests/functional/test_recipes.py.
If you change the scope of the
test_client fixture to a 'function' scope:
test_client fixture will run prior to each of the two _with_fixture tests:
(venv)$ pytest --setup-show tests/functional/test_recipes.py ======================================= test session starts ====================================== tests/functional/test_recipes.py ... SETUP F test_client functional/test_recipes.py::test_home_page_with_fixture (fixtures used: test_client). TEARDOWN F test_client SETUP F test_client functional/test_recipes.py::test_home_page_post_with_fixture (fixtures used: test_client). TEARDOWN F test_client ======================================== 4 passed in 0.21s =======================================
Since we want the
test_client fixture to only be run once in this module, revert the scope back to 'module'.
When developing tests, it's nice to get an understanding of how much of the source code is actually tested. This concept is known as code coverage.
I need to be very clear that having a set of tests that covers 100% of the source code is by no means an indicator that the code is properly tested.
This metric means that there are a lot of tests and a lot of effort has been put into developing the tests. The quality of the tests still needs to be checked by code inspection.
The other extreme where this is a minimal set (or none!) of tests is a very bad indicator as well.
I recommend using pytest-cov based on its seamless integration with pytest. It's built on top of coverage.py, from Ned Batchelder, which is the standard in code coverage for Python.
Running pytest when checking for code coverage requires the
--cov argument to indicate which Python package (
project in the Flask project structure) to check the coverage of:
(venv)$ pytest --cov=project ============================= test session starts ============================== tests/functional/test_recipes.py .... [ 30%] tests/functional/test_users.py ..... [ 69%] tests/unit/test_models.py .... [100%] ---------- coverage: platform darwin, python 3.8.5-final-0 ----------- Name Stmts Miss Cover ------------------------------------------------- project/__init__.py 27 0 100% project/models.py 32 2 94% project/recipes/__init__.py 3 0 100% project/recipes/routes.py 5 0 100% project/users/__init__.py 3 0 100% project/users/forms.py 18 1 94% project/users/routes.py 50 4 92% ------------------------------------------------- TOTAL 138 7 95% ============================== 13 passed in 0.86s ==============================
Even when checking code coverage, arguments can still be passed to pytest:
(venv)$ pytest --setup-show --cov=project
This article provides a guide for testing Flask applications, focusing on:
- Why you should write tests
- What you should test
- How to write unit and functional tests
- How to run tests using pytest
- How to create fixtures to initialize the state for test functions
If you're interested in learning more about Flask, check out my course on how to build, test, and deploy a Flask application: