dockerfixtures 0.2.4

Cointainers as Pytest fixtures made easy

This package was inspired by others, but after trying to make heads and tails of them when faced with bugs, I ended-up finding simpler to start over.

Using dockerfixtures with pytest

To spawn a container in your tests, proceed as follow:

import docker
from dockerfixtures import image, container
import pytest

@pytest.fixture(scope='session')
def docker_client():
     return docker.from_env()

@pytest.fixture(scope='session')
def pg_image() -> image.Image:
    return image.Image('postgres', tags='12')

@pytest.fixture(scope='function')
def pg_container(docker_client: docker.client.DockerClient,
                 pg_image: image.Image) -> container.Container:
    yield from container.fixture(docker_client, some_image)

# If you don't need to reuse the image

@pytest.fixture(scope='session')
def pg_container(docker_client) -> container.Container:
    some_image = image.Image('postgres', tags='12')
    yield from container.fixture(docker_client, some_image)

Why not a pytest plugin ?

Other implementation of this have been provinding a pytest plugin, so you might wonder why this one doesn’t ?

First reason is I have not looked into it that much, yet.

But anyhow, you would still need to import the dockerfixtures.image module. So I am not very sure what the benefits would be ?

Also I found those plugins to provide somewhat bizarre API, for example to define the fixtures’ scope. I haven’t looked into why they do that, yet. Here there are no surprises, a container fixture looks like any other fixture.

Pytest plugins are global: they have to be imported in your top-level conftest.py (see note). I think it is good practice to keep your tests properly partitioned based on their external dependencies. It can help split workload if the need arises. In a collaborative environment, having to import dockerfixtures, may help prevent breaking that partitioning during reviews.