Open johncoxon opened 1 year ago
Thanks for the suggestion!
We've tried distributing through both PyPI and conda, but the support overhead with conda was too much.
Can you clarify how it would be more convenient? Is there a situation where you can't install PySPEDAS in your conda environment using PyPI?
It means that if I want to add packages to an environment that I've installed PySPEDAS in, I have to create an entirely new environment because pip
makes changes to the environment that conda
is unaware of.
It also means, for instance, that I can't give students a one-line method of creating a conda environment that includes PySPEDAS, which means I'm less likely to use it when supervising summer projects or master's projects, reducing its utility to me and presumably others in the community.
Would it help to give them a conda environment.yaml? That's how I've been installing it - everything conda has gets managed by conda, everything it doesn't have gets put in the pip section. I can upload it as soon as I've run it from scratch and made sure it isn't stoopid.
This .yml has conda install everything that can be conda installed - including dependencies of the things that can't be conda installed, like bokeh (which is no longer listed as a pyspedas requirement), matplotlib, and scipy - and the only things left to be grabbed from PyPI are
Since all of the dependencies of (dependencies of (...)) things that can be found in base or conda-forge are kept in conda's box, I've been able to install other things to such an environment without problems - or, maybe stated better, nothing else I've installed into this environment has needed these dependencies and none of their dependencies have been borked by pyspedas.
Rename me to pyspedas.yml, install with conda env create -f pyspedas.yml
Thank you, this is fabulous in the interim!
It would be very useful to be able to install pyspedas using
conda
, since it would make using this framework much more convenient in conda environments.