/docker-ijulia

Docker container with Jupyter and IJulia configured

Primary LanguageJuliaMIT LicenseMIT

Jupyter docker image with Julia kernel

This image contains Jupyter and Julia language kernel based on a fork of IJulia. It uses phusion/baseimage as a base image.

Currently two Julia packages are customized:

  • ivant/IJulia.jl, which fixes package build to make it put the kernel spec to .local/share/jupyter/kernels.
  • ivant/Bokeh.jl, which fixes some compilation issue (probably due to changes in type inference engine).

Usage

Run the container:

docker run -d -p 8888:8888 b.gcr.io/ivant-docker-images/juputer-ijulia

and open localhost:8888 in your browser.

Note: Jupyter in this image is not password protected, nor does it use https, and anyone can access the notebook UI.

Known issues

  1. Due to ongoing changes in package precompilation, many packages fail to load, see issue 12508 of JuliaLang/julia repo. A workaround is to add

    VERSION >= v"0.4.0-dev+6641" && __precompile__(true)

    statement at the beginning of a Foo.jl file for each package Foo that does not already forbid precompilation using __precompile__(false).

    This workaround is automatically applied during Docker build as a short-term solution for the preinstalled packages.

    Should you need to fix the packages you have installed, run the following in your IJulia session:

    run(`/root/fix_package_precompilation.sh`)