This repository contains the code from Schöley, J., 2021. The centered ternary balance scheme: A technique to visualize surfaces of unbalanced three-part compositions. Demographic Research 44, 443–458. https://doi.org/10.4054/DemRes.2021.44.19 . Compared to the code that is published along with the original article, the code in this repository is containerized to make it work as of 2025-03-13, as the code relies on deprecated R packages that cannot be easily installed anymore.

In case the GitHub container registry starts charging for storate of containers or becomes unavailable in the future, the preserved container images for both Docker and Apptainer are available at: .

Below is a very brief log of actions performed to inspect the code, identify and lock the required R packages, build the container and push it to the GitHub container registry so that it is available to the mybinder.org service for running in the cloud with just a web browser. This log is not intended for novice users and is more of a scratchpad used by the author.

docker run --platform linux/amd64 -it --rm -v "$(pwd):/home/rstudio" -p 8787:8787 rocker/geospatial:4.2.2

install.packages("renv", repos = "https://p3m.dev/cran/__linux__/jammy/latest")

renv::init(force=TRUE)

Restart R

renv::settings$ppm.enabled(value = TRUE)
renv::settings$snapshot.type(value = "all")
options("renv.config.pak.enabled" = TRUE)

Also add options("renv.config.pak.enabled" = TRUE) to .Rprofile.

install.packages("pak", repos = "https://p3m.dev/cran/__linux__/jammy/latest")

renv::status()

Many packages will still be missing, install them manually as you go through the R scripts. Easiest way to install them is to open the source files and let RStudio show the button with missing dependencies and click install. The background job will use renv and pak. Some packages will be missing even after that, so install them manually.

renv::snapshot()

install.packages(c("cpp11", "progress", "RcppArmadillo"))

renv::snapshot()

renv::status()

Stop the docker container.

Create the Dockerfile in Dockerfile4build.

docker build --platform linux/amd64 -f Dockerfile4build/Dockerfile -t r422scholey2021 .

docker tag r422scholey2021 ghcr.io/e-kotov/demographic-research.44-19-containerized:latest

docker push ghcr.io/e-kotov/demographic-research.44-19-containerized:latest

Create a Dockerfile file for Binder in the root of the project.

Dump the container image to a tar file:

docker save -o ~/Downloads/r422scholey2021.tar r422scholey2021

In GWDG HPC (or other HPC, or other Linux system that you have access to), build an Apptainer container image from the Docker container image. Steps may vary, we use GWDG HPC as an example.

Start an interactive job on the HPC:

srun --partition jupyter --pty bash

Switch to the directory with the Docker tar file container image:

cd ~/scratch/containers/

Enable the Apptainer:

module load apptainer

Assuming the Docker image is saved as r422scholey2021.tar and uploaded to ~/scratch/containers/ and you are already in this folder on the HPC:

apptainer build r422scholey2021.sif docker-archive:r422scholey2021.tar

Alternatively, if the container is still available from the GitHub Container Registry, the Apptainer image can also be built directly from there:

apptainer build r422scholey2021.sif docker://ghcr.io/e-kotov/demographic-research.44-19-containerized:latest

Wait for:

INFO:    Creating SIF file...
INFO:    Build complete: r422scholey2021.sif

When done, check that the container image is there:

find r422scholey2021.sif

Now it can be downloaded from the HPC, and re-uploaded for preservation to Zenodo along with the Docker tar image.