Improve mybinder

[eliselavy]

Can you find better way with R's kernel : https://github.com/C2DH/template_repo_JDH_R as we are still with an old version of R:

R version 3.6.3

should be version 4.4.3

---

GitHub action to check the laucnh on myBinder

---

Create a branch for the template repo https://github.com/C2DH/template_repo_JDH with the new syntax for mybinder with dockerfile

https://mybinder.readthedocs.io/en/latest/tutorials/dockerfile.html

You can use this repo as an example:

https://github.com/jdh-observer/aADj3Ljjqti2

---

[eliselavy]

@memerchik now you find a docker image allowing to run the version 4.4.3, can you test the docker image locally with your changes
https://github.com/C2DH/journal-of-digital-history-jupyter-stack

https://github.com/C2DH/journal-of-digital-history-jupyter-stack?tab=readme-ov-file#for-rs-user

[memerchik]

C2DH/template_repo_JDH_R#1

Information about the jupyter/r-notebook:x86_64-r-4.3.1 image

1. Core Stack Foundation

r-notebook inherits from jupyter/minimal-notebook → jupyter/base-notebook → jupyter/docker-stacks-foundation. That gives:

• An unprivileged user jovyan (UID 1000) owning /home/jovyan & /opt/conda.
• Package managers conda & mamba.
• A tiny init wrapper (tini) plus a start.sh entrypoint that correctly launches the Jupyter server

2. R & Notebook Components

• R runtime & kernel
  -- R 4.3.1 interpreter installed via conda-forge.
  -- IRkernel already registered, so you get an “R” kernel out of the box.
• Tidyverse bundle
  -- Popular R packages—ggplot2, dplyr, tidyr, readr, purrr, stringr, forcats, etc.—are pre-installed from conda-forge

3. Default Behavior & Directories

Working directory

Containers start with WORKDIR=/home/jovyan and Binder (or your own Dockerfile) usually copies notebooks into /home/jovyan/work.

Entrypoint & port
The stock CMD ["start.sh"] will launch jupyterhub-singleuser (or JupyterLab/Notebook) on port 8888, with token-based auth by default.

4. Extending the Image

Adding R packages

Use a simple install.R (or renv::restore()) and run it in your Dockerfile; it installs into ~/.local/lib/R for jovyan.

System libraries

If you need OS-level deps (e.g. libxml2-dev, libfontconfig1), supply an apt.txt file (one package per line) rather than editing apt-get commands directly in your Dockerfile—this keeps layer caching predictable.

@eliselavy

[memerchik]

Here is a template for the Python-based jupyter notebook using a Dockerfile instead of the requirements.txt and runtime.txt files. In this template, the Python version as well as dependencies & their versions are specified inside of the Dockerfile.

https://github.com/memerchik/template_repo_JDH/tree/dockerfile

@eliselavy

[eliselavy]

@memerchik please integrate the change made for R template here https://github.com/C2DH/journal-of-digital-history-jupyter-stack?tab=readme-ov-file#for-rs-user