mfansler
commented
1 year ago I cannot recreate this issue, at least how it is narrated. E.g.,
CONDA_SUBDIR=linux-aarch64 mamba create -dn foo -c conda-forge r-terra python=3.10solves without issue, implying that the blame is not correctly assigned. Please provide reproducible commands to encounter the issue. For example, a YAML would be preferable.
More importantly though: this just is not a good setup. If Python and R are getting co-installed in one environment, there is almost always a more effective and maintainable configuration. At a glance this should be at least three environments:
- Jupyter infrastructure
- Python kernel
- R kernel
The entry point for the container would be the Jupyter environment, which should have nb_conda_kernels installed. The other two environments should have their respective kernel packages, so they will be available through Jupyter, namely ipykernel and r-irkernel, respectively.
nathancarter
Solution to issue cannot be found in the documentation.
Issue
If you try to
conda install -c conda-forge r-terrain a Python 3.10.5 context, it fails with anUnsatisfiableErrorbecause some of the packages on whichr-terradepends do not support Python 3.10.5 (e.g.,asttokens). In fact, it seems thatasttokensrequires something strictly earlier than Python 3.9. (This is not necessarily the only culprit dependency; it is just the first one in alphabetical order that I noticed.)My use case is that I'm building a Docker container based on
jupyter/datascience-notebookand want to installr-terraas part of the container build process. I'm using an M1 (ARM) machine, so I'm specifying a base image with a tag,jupyter/datascience-notebook:aarch64-SOMETHING, such asaarch64-ubuntu-20.04, to ensure the container works on my architecture. But all ARM base images are from recent years (not a surprise) and thus include recent versions of Python, not Python 3.8.*. So I am unable, in this particular use case, to simply downgrade my Python.Short version:
r-terraneeds old Python, M1 Mac needs new Python, so they are not friends.I could perhaps install multiple Python versions in the container, but I do not know whether that would solve this headache or cause new ones.
Installed packages
Environment info