initial config without riotiler

[emmalu]

rio-tiler and rioxarray packages both give the following:

ERROR: A GDAL API version must be specified. Provide a path to gdal-config using a GDAL_CONFIG environment variable or use a GDAL_VERSION environment variable.

[emmalu]

@wildintellect one thing that comes to mind is: should we 'mix' the use of both pip and conda to handle the different package installs? or perhaps move everything to be handled by conda?

[wildintellect]

Generally move as much as possible to conda, I think is the right answer. Only resort to pip when conda doesn't have the right versions. But I'll look at this more closely.

[wildintellect]

The bigger issue is that the pip is being run outside the conda environment. I dug deeper...

• https://github.com/devcontainers/images/blob/794860b0aad041def006226c0849c1096a1edf8d/src/miniconda/.devcontainer/Dockerfile#L46-L48 does not read the environment.yml (this is what builds the microsoft image `mcr.microsoft.com/devcontainers/miniconda:0-3) so that all happens before the image is pulled. "For convenience, this definition will automatically install dependencies from the environment.yml" appears to be untrue, source, I don't think they use the right Docker command. I think it's supposed to be ONBUILD RUN not RUN on https://github.com/devcontainers/images/blob/794860b0aad041def006226c0849c1096a1edf8d/src/miniconda/.devcontainer/Dockerfile#L47 🐞 Pangeo has a good example https://github.com/pangeo-data/pangeo-docker-images/blob/master/base-image/Dockerfile
• Enabling those lines in the Dockerfile fails, because it's executing as the wrong user vscode instead of root. I'm actually thinking this whole image is useless, even though it gives the conda group ownership over the conda install, the permissions are not correct all files and attempting to install anything with conda fails as the the vscode user. We probably should report a 🐞
• pip install once the container is up and at a conda terminal does work, no issues. I could load the widget in the VSCode notebook viewer but it doesn't work right with ipywidgets. So we do need jupyterlab, @abarciauskas-bgse the readme doesn't have any instructions on if anything is needed to register ipyleafet inside jupyterlab.

Options

• Start with a different base image (we could reuse some of the method MS uses) or go the micromamba route (side note: this version of miniconda does not seem to have the libmamba solver - no idea why)
  • https://github.com/mamba-org/micromamba-docker
• Try changing to root to add to the conda install, or change the permissions to make it all owned by vscode user. Technically worked but gave me some grief. @emmalu do you want a PR to your PR?

USER root
COPY environment.yml* .devcontainer/noop.txt /tmp/conda-tmp/
RUN if [ -f "/tmp/conda-tmp/environment.yml" ]; then umask 0002 && /opt/conda/bin/conda env update -n base -f /tmp/conda-tmp/environment.yml; fi \
    && rm -rf /tmp/conda-tmp

[wildintellect]

@emmalu I think I have a fix, see #49

[wildintellect]

If this is working now for @emmalu and @edkeeble then please proceed.

[review-notebook-app]

Check out this pull request on 

See visual diffs & provide feedback on Jupyter Notebooks.

---

Powered by ReviewNB