pytest-recording is a neat pytest plugin that makes it easy to use the VCR library, which helps write tests against HTTP resources by automatically capturing responses and baking them into a YAML file to be replayed during the tests.
It even works with boto3!
To use it, first install it with pip install pytest-recording
and then add the @pytest.mark.vcr
decorator to a test that makes HTTP calls:
@pytest.mark.vcr
def test_create():
runner = CliRunner()
with runner.isolated_filesystem():
result = runner.invoke(cli, ["create", "pytest-bucket-simonw-1", "-c"])
assert result.exit_code == 0
The first time you run the tests, use the --record-mode=once
option:
python -m pytest -k test_create --record-mode=once
This defaults to creating a YAML file in tests/cassettes/test_s3_credentials/test_create.yaml
.
Subsequent runs of pytest -k test_create
will reuse those recorded HTTP requests and will not make any network requests - I confirmed this by turning off my laptop's WiFi.
For my tests that use API key services like the OpenAI API I've settled on this pattern:
import os
import pytest
API_KEY = os.environ.get("PYTEST_OPENAI_API_KEY") or "fake-key"
@pytest.fixture(scope="module")
def vcr_config():
return {"filter_headers": ["authorization"]}
The vcr_config()
there ensures that any authorization
HTTP headers in the resulting cassette YAML files are redacted, so they don't accidentally get checked in to a public GitHub repo.
I use that PYTEST_OPENAI_API_KEY
environment variable in tests like this:
@pytest.mark.vcr
def test_ask(monkeypatch):
monkeypatch.setenv("OPENAI_API_KEY", API_KEY)
# ...
Then for the duration of that test the OPENAI_API_KEY
environment key will either be fake-key
for test runs against previously recorded cassette or the value of PYTEST_OPENAI_API_KEY
- which means I can update the cassettes like this:
PYTEST_OPENAI_API_KEY=your-key python -m pytest --record-mode once
Created 2021-11-02T17:51:51-07:00, updated 2024-11-24T12:56:52-08:00 · History · Edit