cacrawford/jupyter repository overview

⁠Docker container for Jupyter with Pipenv

For those that need a Docker instance for running local Jupyter servers. Useful if you have sensitive data on your machine that only you need to access, and you have comfort using docker-compose⁠. Also, since the notebook is containerized and only installs dependencies from your Pipfile, is a great way to track what dependencies are being used from source control.

I'm working on making the Jupyter container as configurable as possible.

NOTE: This container does change your Pipfile! It will install ipykernel (as a dev dependency) in order to link the kernel to Jupyter.

⁠Setup

Optionally, you should have pipenv⁠ installed.

Initialize a new project in your desired directory with

pipenv --three

You will also need to export your UID (on Mac) or assign your UID to your user ID (Linux) in order for the container to write to your local volume as a non-root user. On Mac, this is:

export UID

Then, add a docker-compose.yml with the following service:

jupyter:
  image: cacrawford/jupyter
  volumes:
    - .:/ws
  user: $UID
  environment:
    ## needs to be a hashed password
    ## generate with (assuming jupyter is installed)
    ##   from notebook.auth import passwd()
    ##   passwd()
    ## if left empty, the notebook will generate a token which you can find with
    ##   docker-compose logs -f jupyter
    - NOTEBOOK_PASSWORD=YOUR_HASHED_PASSWORD
  ports:
    - "127.0.0.1:8888:8888"           # restricts this notebook to the host machine

If your Pipfile has a lot of dependencies, this may take a bit to start up.

⁠Usage

Make sure that any notebook you have uses the python3-pipenv kernel, otherwise your dependencies won't be included.

In order to install new dependencies on the container, run

docker-compose exec jupyter pipenv install <packages>

This will update your local pipenv file but only install packages on the container machine.