Open sterlinm opened 1 year ago
I can look into this. It would be a major change in behavior, but most users that assume it pulls in R will almost certainly also be installing other R packages. I think it makes sense to decouple them and would make the package much more flexible. Also cut down this [2 x 4 x 6] tensor of variants.
Thanks! I understand that it might be a big change, but agree that it seems like there probably aren't many people who are relying on including rpy2 in their environment to implicitly include r-base and other packages.
Right now, if you want to use rpy2 in a conda environment with a separately managed installation of R it can be challenging because you end up with r-base on your path as well or you have to create a broken environment by installing with --no-deps.
I can try to put together a PR if you're open to the idea.
Maybe a multiple output recipe where the rpy2 keeps the status quo and a rpy2-base without r-base?
I don't think there's any reason that r-base needs to be a requirement for installing rpy2.
If you want to install R independent of conda or in a separate conda environment, automatically including r-base and its dependencies in the environment with rpy2 just makes it more difficult to point it to the version of R you want.
Would you consider moving r-base from host to the test requirements section?
Apologies if I missed anything. Happy to do a pull request.