Record of work on py-rocket-base and py-rocket-geospatial

[eeholmes]

Design intent Follow Pangeo for python and Rocker for R structure. As you move between the JupyterLab, VSCode, RStudio browser tabs, the environment should feel familiar. You can make good guesses about the structure of the environments if you have worked with pangeo or rocker images before.

Note I am not using the development geospatial script from Carl and some new spatial features might not work.

I initially started with Carls' images here
https://github.com/eeholmes/earthdata-cloud-cookbook/tree/main/.devcontainer

This proved an unstable base image and I wanted something closer to openscapes/py-rocket which follows the Pangeo structure. The Pangeo image follows repo2docker design in that the presence of certain files in the repo, triggers behaviors when you build the image. The openscapes/py-rocket image doesn't do that but does have the structure of a conda environment (notebook) that is activated automatically.

I found the venv mixed with conda, hard to deal with. So went with pure conda/mamba like openscapes/corn and py-rocket.

However the images use the rocker base image rocker/verse and the rocker script for creating the geospatial environment.
https://github.com/rocker-org/rocker-versioned2/tree/master

Much of the work was on getting the base to start up with a

docker run --platform linux/amd64 -p 8888:8888 ghcr.io/nmfs-opensci/container-images/openscapes-rocker:latest

call. I was getting a R not found error when trying to start RStudio.

I needed to add the files from \home\rstudio into \home\jovyan. This is wiped out in a JHub but needed if run with docker run. Also it seemed like it wanted the notebook module.

RUN usermod -aG sudo ${NB_USER} && chown -R ${NB_USER}:${NB_USER} /srv
RUN mkdir /home/jovyan \
    && usermod -d /home/jovyan rstudio \
    && chown -R "${DEFAULT_USER}:${DEFAULT_USER}" "/home/jovyan" \
    && cp -r /home/rstudio /home/jovyan