SageMath Docker Image

This is an attempt to standardize and streamline computational work that combines SnapPy, SageMath, and friends via a custom Docker image that is based around Ubuntu 16.04 LTS and the latest and greatest SageMath. It is indirectly derived from the official sagemath/sagemath Docker image, but includes SnapPy, and the Python interfaces to Regina, and PHCPack.

A container instance of this image starts a shell as the user sage with all the environment variables set as if you had executed sage -sh. In particular, python is SageMath's Python.

This image takes a long time to build, even on a fast computer, since it compiles SageMath from source, and weighs in at just under 4GB (1GB compressed). We recommend you use the posted image on DockerHub via:

docker pull computop/sage
docker run -it computop/sage

Here, the first command downloads the image from DockerHub and the second starts a container that is running said image. Moving files between the container and the rest of the world can be done using scp or having the container mount a local directory as in this example:

docker run -it --mount type=bind,source="$HOME/Dropbox/linux_share",target=/home/sage/linux_share computop/sage

where the directory linux_share inside the user's Dropbox folder is shared.

If you want to use Sage's Jupyter notebook interface, start the container via:

docker run -it -p computop/sage

and in said container type:

sage --notebook

and then point your computer's web browser to http://localhost:8888. If you want to use the beta JupyterLab, which will eventually superseded Jupyter, you can instead do:

sage --notebook jupyterlab

In JupyterLab, you have access not just to notebooks but also Python consoles and shell terminal windows.

Other uses

SageMath and all the extras are installed in /sage. You can make a tarball of that and copy it over to another Ubuntu 16.04 machine. You will need to have certain system packages installed for this to work. Those listed here are certainly enough, though surely one could get by with much less.


A partial list of included software:


To build from scratch, which takes 1.3 hours using all 8 cores on a Mac Pro:

docker build --tag=computop/sage sage

If compiling Sage fails, make sure the Docker application is configured so that it can use at least 4GB of memory.

Put out on DockerHub:

docker push computop/sage