Closed corneliusroemer closed 1 year ago
Pending discussion of if it's ok to remove Pangolin/pangoLEARN or not in https://github.com/nextstrain/conda-base/issues/7.
Let's zap pangolin and pangolearn here - they are so out of date, they aren't useful at all. And they constrain our snakemake version to be mismatched with docker-base.
@corneliusroemer I don't appreciate that you merged this unilaterally without addressing anything in https://github.com/nextstrain/conda-base/issues/7 and with a sparse commit message here https://github.com/nextstrain/conda-base/commit/560e502589bf237eb66b94428f58e8b9528497a4 that's mostly noise and leaves a lot to be desired.
Even if we're going to dismiss the potential issues of changes—and I do think that's justifiable in this case—I think it's still important to at least address them.
Description of proposed changes
pangolin and pangolearn have some tricky dependency pins that create issues. This is a known problem.
As a result, I set up separate environments for pangolin/pangolearn/usher, different from the rest of the stack.
pangolin/pangolearn is not used by any Nextstrain workflows as I'm aware, so we don't need to include them. Nextclade does the job just fine. Pangolearn now seems to require so much memory that it's not runnable on normal consumer hardware anyways (>32GB RAM).
Hence I propose to remove pangolin/pangolearn. I also pin iqtree >=2 and add some version requirements for augur/auspice/nextstrain-cli to save us from accidental downgrades when adding a new package.
Resolves #7