experimental: adding support for wheel packags to repodata parsing

[travishathaway]

Description

This pull request is directly related to this work:

• Discovery: early testing of rattler-solver with repodata v3 conda-incubator/conda-pypi#168

Please consider this just one idea for how it could be implemented, and a way to start a conversation. This is not meant to be merged in its current state.

[baszalmstra]

Thanks for taking the time to contribute this. Overall this particular part looks fine to me!

But before merging this though I would like to understand much better what this does. Is there a CEP (even draft) that goes over the details?

Are the filenames still conda filenames? They follow a strict pattern that is relied on in this code.

Can you also add some test repodata to both show what it would look like and to have some form of regression data.

[travishathaway]

Hi @baszalmstra,

Thanks for the quick response. Just to give you some more background on this, we are currently implementing a conda plugin that will allow us to install wheel files by performing a solve against a repodata.json file that contains a top-level packages.whl key. For more information about how this repodata file will look, you can check out an example here:

• Discovery: early testing of rattler-solver with repodata v3 conda-incubator/conda-pypi#168 (comment)

Are the filenames still conda filenames? They follow a strict pattern that is relied on in this code.

No, they are not. They are .whl files. I'll be happy to add necessary edits elsewhere in the repository to make sure it's fully supported. As well as adding the extra tests too.

Yes, a CEP would be great here to define the exact format of this new section option section in repodata.json. I would be happy to write this.

[travishathaway]

Hey @baszalmstra,

I just wanted to post this here for visibility:

• Add repodata support for filenames in the wheel format conda-incubator/conda-pypi#190 (comment)

This pull request is far from complete (i.e. the many TODOs I had to add today 🙈), but it's nice to know that it kind of works! 😅

The thing that makes me the uneasiest about it is the way I setup repodata file type merging. I think it's okay to assume that the .conda file type is always going to be the preferred file type to choose by default. I'm just not sure I've handled the both case correctly. Now, that there's three, it also feels weird calling it "both".

Anyways, these are things to be spelled out and debated in a proper CEP 🤝 🔏 , but I just wanted to give my thoughts on it so far here too.