Local development with Docker

Docker Compose is the only officially supported development configuration. We’re doing this to stay focused on the project goals rather than triage lots of “not working on my machine” issues. If you wish to use virtualenvs instead, please ensure that you run all of the tests with Docker Compose before opening any Pull Requests.

The steps below will get you up and running with a local development environment.

Note: All of these commands assume you are in the root of your generated project.


Install Docker

You’ll need a recent version of Docker. If you don’t already have it installed, follow the instructions for your OS:

Docker for Mac and Windows include Docker Compose, so most Mac and Windows users do not need to install it separately. For Linux users, follow the instructions for Linux installation in the Docker documentation

Build the Stack

This can take a while, especially the first time you run this particular command on your development system:

$ docker-compose -f local.yml build

Boot the System

This brings up the interface app (running Django), prediction app (running Flask), and PostgreSQL.

The first time it is run it might take a while to get started, but subsequent runs will occur more quickly.

Open a terminal at the project root and run the following for local development:

$ docker-compose -f local.yml up

You can also set the environment variable COMPOSE_FILE pointing to local.yml like this:

$ export COMPOSE_FILE=local.yml

And then run:

$ docker-compose up

You should now be able to open a browser and interact with the services:

  • interface frontend at http://localhost:8080
  • interface API at http://localhost:8000/api/
  • prediction service at http://localhost:8001
  • documentation service at http://localhost:8002

Run tests

For the vue frontend interface, you can run the unit test with:

$ docker-compose -f local.yml run vue_unit_test

And the end to end test with:

$ docker-compose -f local.yml run vue_e2e_test

Code changes

Because local.yml has the volume mappings ./interface/:/app and ./prediction/:/app (the project dirs containing all of the code into the container’s source code directory), changes made to files during development should be reflected in the running interface or prediction container as soon as the dev server process restarts.

Create JSON data

First use the testing fixture factories to create some data in your local database:

docker-compose -f local.yml run interface python manage.py shell_plus

Python 3.6.2 (default, Sep  8 2017, 06:15:13)
[GCC 4.9.2] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> from backend.cases.factories import *
>>> case = CaseFactory()
>>> candidates = CandidateFactory.create_batch(5, case=case)
>>> nodule0 = NoduleFactory(candidate=candidates[0])
>>> nodule1 = NoduleFactory(candidate=candidates[1])
>>> nodule4 = NoduleFactory(candidate=candidates[4])

Then run the API: docker-compose -f local.yml up interface

And if you go to http://[IP of your docker, could be localhost]:8000/api/ you can play with the data you just created:

For instance, here’s the list of nodules we just created:

        "url": "http://localhost:8000/api/nodules/1/",
        "centroid": {
            "id": 6,
            "x": 324,
            "y": 285,
            "z": 59,
            "series": 6
        "created": "2017-10-10T22:16:35.620318Z",
        "lung_orientation": 0,
        "case": "http://localhost:8000/api/cases/2/",
        "candidate": "http://localhost:8000/api/candidates/1/"
        "url": "http://localhost:8000/api/nodules/2/",
        "centroid": {
            "id": 7,
            "x": 49,
            "y": 227,
            "z": 31,
            "series": 7
        "created": "2017-10-10T22:16:44.436150Z",
        "lung_orientation": 0,
        "case": "http://localhost:8000/api/cases/3/",
        "candidate": "http://localhost:8000/api/candidates/2/"
        "url": "http://localhost:8000/api/nodules/3/",
        "centroid": {
            "id": 8,
            "x": 242,
            "y": 339,
            "z": 22,
            "series": 8
        "created": "2017-10-10T22:16:50.403528Z",
        "lung_orientation": 0,
        "case": "http://localhost:8000/api/cases/4/",
        "candidate": "http://localhost:8000/api/candidates/5/"

Running the tests

The tests/test_docker.sh shows you how to run the tests for both the prediciton and interface applications.

For example, if you are actively developing the prediction app, you could run the following to execute the tests:

$ docker-compose -f local.yml run prediction pytest

Other notes

Pre-commit hooks

Git pre-commit hooks are useful tools that run commands before you commit. To enable easier maintainance of style guide and running tests, we implement pre-commit git hook commands that will run every time before you create a commit.

To set it up on your local development repository:

$ cd concept-to-clinic

Copy the .githook-pre-commit file into the .git/hooks/ folder as pre-commit with -f flag so that you overwrite any existing pre-commit file:

$ cp -f .githook-pre-commit .git/hooks/pre-commit

Install flake8 locally for the styling scripts:

$ pip install flake8 pep8

You are set! The next time you want to commit, styling and testing pre-commit commands will be executed first and if they fail, your commit will not be committed.

If you want to by-pass the pre-commit check, use --no-verify flag in your commit:

$ git commit --no-verify -m "Ignore pre-commit rules"

To check if the created pre-commit hook is working without having to do a commit, you can run from the root of the project:

$ .git/hooks/pre-commit

Detached Mode

If you want to run the stack in detached mode (in the background without console output), use the -d argument:

$ docker-compose -f local.yml up -d