Can we autogenerate the catalog yaml (and maybe a prefilled recipe)?

[jbusecke]

Looking at the template catalog.yaml file I suspect that there will be issues with people just not editing the paths and then stores get overwritten...

Wondering if there is an easy way to generate these paths (and maybe some other dummy content) for the catalog.yaml. The template for that would be something like:

# All the information important to cataloging.
"ncviewjs:meta_yaml_url": "https://github.com/<owner>/<repo_name>/blob/main/feedstock/meta.yaml"
tags:
  - zarr
stores:
  - id: <recipe_id (parse_from_meta_yaml)>
    url: "gs://leap-scratch/data-library/feedstocks/latest/<repo_name>/<recipe_id>.zarr"

Not super important rn, but just putting this out there. Maybe @andersy005 has an idea?

[jbusecke]

Thinking more about this. These sort of patterns seem bad (and maybe insecure?).

I am tending more towards the following now:

• Have a very thin meta.yaml file and some default requirements.txt in the template
• query the user to fill that out
• run an 'init_feedstock.pyscript, that sets up a reasonable recipe template, catalog.yaml based on the info we can get from the repo? (This would not work with the Juptyer labusername, but we could just handle that with adata-library-local-test....` prefix, since the chances that two users work on the same feedstock are low?

[jbusecke]

This could also take care of avoidable errors like I did here (jobnames cannot contain underscores). EDIT: This is only relevant to Dataflow, so we should probably handle this on the action level