test: add tests on handling of duplicate matrix selectors

[jameslamb]

Contributes to rapidsai/build-planning#31.

Over in rapidsai/devcontainers#365, I'm proposing a change that depends on how rapids-dependency-file-generator handles duplicates in --matrix selectors.

See rapidsai/devcontainers#365 (comment).

This proposes adding tests covering that behavior, to reduce the risk of accidentally removing that support in future refactorings.

[KyleFromNVIDIA]

This is fine, but if it was me, I would have modified rapidsai/devcontainers#365 (comment) to detect if the user is overriding cuda_suffixed to avoid passing more than one value.

Let's explicitly document this behavior too.

[jameslamb]

This is fine, but if it was me, I would have modified rapidsai/devcontainers#365 (comment) to detect if the user is overriding cuda_suffixed to avoid passing more than one value.

The simplest implementation I can think of involves tracking the defaults as a bash associative array, splitting each --matrix-entry value by = into (key, value), and then indexing into that associative array with them.

That was more complexity than I was comfortable with for this purpose, but I'm willing to try it. Let's please not hold up rapidsai/devcontainers#365 or this PR for that though. I'd like to try to get rapidsai/devcontainers#365 and the corresponding dependencies.yaml changes out before code freeze for 24.08 (end of day tomorrow).

Let's explicitly document this behavior too.

Document what, where? I don't understand what you're asking for, sorry.

[KyleFromNVIDIA]

Document what, where? I don't understand what you're asking for, sorry.

Document the fact that the last duplicate is the one that DFG actually uses, which is what this PR is all about.

[jameslamb]

Document the fact that the last duplicate is the one that DFG actually uses, which is what this PR is all about.

Sure, but I'm not sure where you were looking for docs to be added (doc strings in code? the README? one of those, but over in the devcontainers code relying on this?).

I pushed cfe81b4 adding a line to the README, is that what you were looking for?

[KyleFromNVIDIA]

The README, because that's what the end user is going to see.

[jameslamb]

Gotchu! Ok yeah no problem, I agree that's a good addition.

Thanks for the quick review, always appreciated!

[GPUtester]

🎉 This PR is included in version 1.15.0 🎉

The release is available on:

• GitHub release
• PyPI release
• Conda release - rapidsai
• Conda release - rapidsai-nightly

Your semantic-release bot 📦🚀