Material for my workshop at EuroPython 2018
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.
Failed to load latest commit information.


Writing and Running Tests in Docker

Material for my workshop at EuroPython 2018.

Every commit of the repository corresponds to a different step of the workshop.

Getting Started

First, go back in time:

git checkout -b workshop setup

You will start to work from this commit, in a new workshop branch. From there, you can have a look into master if you are blocked and don't know what to do next.

Then, install requirements:

  • VirtualBox: to run the workshop inside a virtual machine (Ubuntu),
  • Vagrant: to set-up the virtual machine,

By using VirtualBox, we are sure to all have the same environment. Which will save us time!

Now that you are ready to start, let's create a virtual machine:

vagrant up  # Create the VM
vagrant ssh  # Connect to the VM

Inside the virtual machine, you can now set-up the demo project:

cd /vagrant  # Local repository is shared with the VM in /vagrant
virtualenv -p python3.6 venv
. venv/bin/activate
pip install -r requirements-test.txt

Finally, you can check that everything is working, by running the demo server:

./ demo --port 5000 --db_url "sqlite:////tmp/europython.sqlite"

And access to the demo blog on, using your favorite browser on your local machine.

Writing Tests (with Pytest)

First step of the workshop is to write tests.

If you try to run them now, you will only get failures:

python -m pytest tests/

By looking at error messages, you will see NotImplementedError exceptions. It's because you have to implement yourself:

  • acceptance tests for the demo blog (using Pytest-BDD) in tests/,
  • end-to-end tests for the demo web API (using WebTest) in tests/,
  • Pytest fixtures, in tests/

To implement the tests:

  1. Execute them one by one:
    • for the blog: python -m pytest tests/
    • for the web API: python -m pytest tests/
  2. Read the error message,
  3. Let you guide by the error message to find a solution,
  4. Implement the solution.
  5. Do it again, until you fix all errors.


  • there are factories in europython/ to create objects in database from the tests,
  • test helpers are available in europython/ and europython/,
  • look at Pytest extensions we use in requirements-test.txt: they already provide fixtures, that our own fixtures or tests are depending on.

Useful links:

Automating Tests (with Tox)

Now that our tests are GREEN, an important step is still missing: how to be sure that our project is packaged correctly? Because if you update a broken package to PyPI, people will not be able to use it...

For this purpose, we will use Tox, to run our tests inside a virtual environment, automatically updated every time you change a dependency or the project's packaging. In fact, we will need several environments:

  • one for running the tests,
  • another one for running lint checking with Flake8,
  • and two development environments:
    • the first one to be used locally (e.g., to get auto-completion in your IDE),
    • the second one to be used in Docker later on.
  • as a bonus, you can also create one for checking security issues in dependencies (with Safety).


  • now that we are using Tox, you can move test dependencies from requirements-test.txt into tox.ini,
  • you can get test coverage for free with

Running Tests (in Docker)

Running tests locally is nice. But only when you work alone!

As soon as you work with other persons, using different operating systems, things start to be complicated. People will probably have different versions of system dependencies, and most of the time, installing the needed version will not be straightforward.

That's where Docker comes to the rescue!

With Docker, the only system dependency people will have to install is Docker itself. Period.

For our workshop, we will need two Docker images:

  1. one for running on production,
  2. another one to use for tests, based on the PROD image (to not dupplicate Dockerfiles), and including all test dependencies like Tox, Pytest and others.

Finally, to simulate a real use case, with a project depending on external systems, we will also replace with PostgreSQL the SQLite in-memory database we were using until now. For this purpose, we will set-up our testing stack with Docker Compose.


  • Use Alpine images, for a smaller image footprint.
  • Configure the application container using environment variables (see europython/
  • You will have to:
    • add psycopg2 to the application's dependencies, and install gcc, libc-dev and postgresql-dev in the PROD application container, in order to interact with PostgreSQL,
    • install chromium and chromium-chromedriver in the TEST application container, in order to run the acceptance tests.
  • You already run tests locally with Tox. So do the same inside Docker 😃 Just don't forget to forward environment variables (used to configure containers) to Pytest, with the passenv directive.
  • Run the tests directly inside the application container in order to:
    • share your local code with the container,
    • take advantage of using PDB when debugging tests.

Automating Workflow (with Invoke)

TODO: explain what to do next.

Adding Continuous Integration (on Travis CI)

TODO: explain what to do next.