My Custom DataScience Jupyter Docker Stack

A personal Jupyter Dockerfile based on jupyter/datascience-notebook with some custom packages, extensions and settings. More info in Jupyter Docker Stacks. Try it with binder!

This setup contains everything that includes jupyter/datascience-notebook and the following:

• conda packages:
  • altair
  • altair_saver
  • dash
  • geopandas
  • jupytext
  • nbgrader
  • pandas-profiling
  • plotly
  • rise
  • vega_datasets
  • xgboost
• JupyterLab extensions:
  • Collpasible Headings
  • Drawio
  • Shortcutui
  • System Monitor
  • Table of Contents
  • Theme Darcula
• Custom JupyrLab user settings
  • Dark theme for the terminal
  • More shortcuts:
    • Restart Kernel and Clear All: Alt+O
    • Restart Kernel and Run All: Alt+P

Additionally, it includes packages for save Altair charts objects using selenium method and Firefox webdriver. For example, the following lines save a simple scatter-plot to PNG.

import altair as alt
from vega_datasets import data

chart = alt.Chart(data.cars.url).mark_point().encode(
    x='Horsepower:Q',
    y='Miles_per_Gallon:Q',
    color='Origin:N'
)

chart.save('chart.png', method="selenium", webdriver="firefox")

More info about saving Altair charts here.

Add more features

You can add other packages from conda or pip, even jupyter notebook/lab extensions modifying the Dockerfile.

In order to add more custom settings you should copy necessary folders and files in the .jupyter folder as any other JupyterLab installation.

Docker usage

In order to build the image run in the same directory where Dockerfile is:

docker build -t <YOUR-IMAGE-TAG> .

Then, start a container based on your new image using JupyterLab running:

docker run --name <YOUR-CONTAINER-NAME> -p <LOCAL-PORT>:8888 -v <YOUR-WORK-DIRECTORY>:/home/jovyan/work <YOUR-IMAGE-TAG> start.sh jupyter lab

I usually run my containers without token on my local machine using the following code:

docker run --name <YOUR-CONTAINER-NAME> -p <LOCAL-PORT>:8888 -v <YOUR-WORK-DIRECTORY>:/home/jovyan/work <YOUR-IMAGE-TAG> start.sh jupyter lab --LabApp.token=''

You can stop the container with:

docker stop <YOUR-CONTAINER-NAME>

Finally, when you want to use your container you have to run:

docker start <YOUR-CONTAINER-NAME>

Alternatively, it is possible to use Visual Studio Code with a Docker extension in order to launch your container as development environment and work with Jupyter Notebooks.