The `pytest` Package

References:

Installation

Installing Pytest:

pip install pytest

Configuration

It helps to add a file called “conftest.py” to the root directory of your repository, or to the “test” directory. This helps pytest find the application code and facilitates successful imports.

Another use for the “conftest.py” file is to define some code, like fixtures, which can be shared across multiple test files.

Usage

The pytest package is generally used as a command-line utility for running pre-defined files of “test” code. Follow the “Testing 1,2,3” Exercise to get acclimated with pytest.

Example invocations:

# run all tests:
pytest

# suppress warnings:
pytest --disable-pytest-warnings

# see print statements:
pytest  -s

# run a specific test:
pytest test/my_test.py

# run a specific function in a specific test:
pytest test/my_test.py -k 'test_my_thing'

Expecting Errors

The pytest package can be imported to facilitate assertions that errors will be raised:

# this is an example test file...

import pytest

def test_divide_by_zero():
    with pytest.raises(Exception) as e_info:
        # we expect this code will raise an error (division by zero)
        result = 2 / 0

Fixtures

If there is a time-intensive or memory-intensive operation used by multiple tests, we can use a “fixture” to perform that operation just once. This is a big time saver.

We’ll generally place the fixtures in the same file as the tests that need them, but if multiple test files need to share the same fixture, we’ll move it to the “conftest.py” file:

# this is the "conftest.py" file

import pytest

# EXAMPLE FIXTURES:
# "module" scope fixtures are only executed once, and shared
# if fixtures need to be shared across multiple test files, we define them here

@pytest.fixture(scope="module")
def nlp():
    import spacy
    print("LOADING THE LANGUAGE MODEL...")
    return spacy.load("en_core_web_md")


@pytest.fixture(scope="module")
def parsed_response():
    import requests
    import json
    print("MAKING A NETWORK REQUEST...")
    response = requests.get("https://www.example.com/api/")
    return json.loads(response.text)

After our fixtures are defined, we can pass them into the test functions as parameters. There is no need to import fixtures that have been defined in the “conftest.py” file, as they will be implicitly loaded and available for use.

# this is an example test file...

def test_my_language_model(nlp):
    # do something with the nlp fixture
    # assert something

def test_my_calculations(parsed_response):
    # do something with the parsed_response fixture
    # assert something

Skipping Tests on CI

If there are tests which invoke functions that issue network requests, we’ll sometimes want to prevent that from happening on the CI server.

To skip any test function, mark it to be skipped conditionally, for example if a CI environment variable is set to “true”. This environment variable is set by default on some CI servers, or you could configure your CI server to run tests with CI=true pytest instead of just pytest (see note below):

# this is an example test file...

import os
import pytest

from app.my_file import fetch_data

CI_ENV = bool(os.getenv("CI") == "true")

@pytest.mark.skipif(CI_ENV, reason="avoid issuing HTTP requests on the CI server")
def test_fetch_data():
    data = fetch_data() # makes a network request
    assert isinstance(data, list) # for example

NOTE: if using GitHub actions, you may need to update the “python.yml” GitHub Actions config file so the last line says CI="true" pytest instead of just pytest. Remember to save the file and make a commit before re-pushing your code.