Chalab

Requires:

• docker

For development (running the tests):

• python3.5
• an env with the content of the requirements.txt (pip install -r requirements.txt).
• phantomjs for browser testing. Other drivers (Firefox, Chrome, etc) can be used too.

Structure

• chalab: core of the project,
• landing: the home page app,
• wizard: the wizard app,
• user: user profile app,
• group: the group app,
• bundler: the challenge -> bundles (zipfiles) app,
• instances: the module containing the settings per environment:
  • local: the settings to run django on your local machine,
  • local_docker: the settings to run django in docker on your local machine (preferred way),
  • production.py: settings in production.
• templates: global templates,
• datasets: default datasets from automl.

We use Celery + RabbitMQ to schedule long running tasks like creating the bundles.

Local Dev

The most straightforward way to run chalab is to use docker and docker-compose. It will start the system (database, server, etc) automatically.

Then you can run the tests on your local machine (non-dockerized) to get assistance from your IDE. As of now it's not possible to run the tests inside Docker (Pull-Requests are welcome!).

Start the server:

In the chalab folder,

make dev

It downloads all the containers & dependencies then start the containers. Access the app at http://localhost:8000/. Kill it (ctrl-c) then re-run make dev to rebuild the container and apply db migrations.

Management

• make superuser: when you started the database (make dev does it for you)
• make static: to re-generate the static resources
• make migrations: to generate the django migrations

Datasets

Chalab comes with a few datasets by defaults, we don't store them in the git repository.

• make predload_db: setup the database with datasets and metrics:
  • make dataset: download the default datasets locally and add them to the database,
  • make metrics: generate the default metrics options and add them to the database.

Note that you should these commands only once or you'll have duplicates.

Testing: setup

Install django, selenium & other dependencies locally.

# in your venv:
pip install -r requirements.txt

Run the tests

# while the server is running using `make dev':
make test

# clean the data produced by tests
make clean

• You can override the selenium ENVs used by tests/selen.py to use different drivers, by default, you need phantomjs installed.
• chrome requires the chromedriver to be accessible from your PATH.
• firefox tends to break if you use the last version.
• phantomjs (default) is a safe choice but you won't see the interactions live.

Deployment

Pull the repo, copy the example.env to .env, then adapt it to your system. Then run make prod

• The production app server will be available at 127.0.0.1:8742
• The production static server will be available at 127.0.0.1:8842

Both will be ready to be accessed behind a reverse proxy such as nginx (standard gunicorn setup).