Open CJ-Wright opened 6 years ago
Thanks for bringing this up @CJ-Wright! This is a setup task in my mind, for sure.
In terms of getting the requirements, I don't like the idea do pulling from a named file. We already have a variable called $DOCKER_CONDA_DEPS
. Maybe we should have one or a few called $CONDA_BUILD_DEPS
, $CONDA_DEPS
and $CONDA_TEST_DEPS
. If folks want to read these from a requirments file, that is fine.
Do we want to push after creating the recipe?
Yes, otherwise what is the point. We probably should fork staged recipes if it doesn't exist.
Do we want to run the tests in the recipe?
Yes
Do we want to pull the maintainers/licence info/other info from somewhere?
Yes.
This feature will have a lot of parametrizations associated with it.
Will we have a version/tag/etc up during setup? Do we need these for a staged-recipes PR?
Unfortunately right now you can't assume a version number for setup, but this is information you need for staged recipes
I guess we could a conditional inside the conda_forge
activity that searches for a feedstock. If none exists then we go and make a PR into staged otherwise continue as usual.
Hmm I am -1 on that behaviour. Getting something through staged-recipes is a once-in-a-project activity. Therefore, it is a setup task, not something that should happen if it doesn't exist.
That's fair
If we are going to do conda-forge PRs maybe we should have something that builds the initial PR into staged-recipes.
Pieces
PushTag
so we can pull it from GH?)requirements/build
andrequirements/run
.requirements/test
Questions