fix: Generate valid CRDs for fields with Optional enums and doc-comments

[NickLarsenNZ]

Fixes #1906

Motivation

This is a continuation of #1821

Solution

Replace the OptionalEnum schema transformer with the one from #1839, and adds coverage to the existing test.

In addition, I moved the schema module into a subdirectory with hopes that the other transformers and their tests will be moved into individual modules below that. I am happy to do that in this PR, but decided to wait until you had a chance to look at it.

[doxxx93]

Thanks for working on this!

I think the module restructuring could be split into a separate PR. It would be easier to review if this PR focuses only on fixing #1906, and the refactoring comes as a follow-up.

What do you think?

[codecov]

Codecov Report

❌ Patch coverage is 95.34884% with 2 lines in your changes missing coverage. Please review.
✅ Project coverage is 76.2%. Comparing base (a75b211) to head (6c533bb).

Files with missing lines	Patch %	Lines
kube-core/src/schema/transform_optional_enum.rs	95.2%	2 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##            main   #1907     +/-   ##
=======================================
+ Coverage   76.1%   76.2%   +0.1%     
=======================================
  Files         86      87      +1     
  Lines       8399    8429     +30     
=======================================
+ Hits        6387    6418     +31     
+ Misses      2012    2011      -1     

Files with missing lines	Coverage Δ
kube-core/src/schema/mod.rs	95.9% <100.0%> (ø)
kube-derive/tests/crd_schema_test.rs	100.0% <ø> (ø)
kube-core/src/schema/transform_optional_enum.rs	95.2% <95.2%> (ø)

... and 2 files with indirect coverage changes

🚀 New features to boost your workflow:

• ❄️ Test Analytics: Detect flaky tests, report on failures, and find test suite problems.

[doxxx93]

I've submitted #1908 with a minimal fix.

Your PR introduces 248 new lines with module restructuring and LazyLock, but the root cause is simpler - the o.len() == 1 check skips schemas when doc comments add a description field.

Just removing that check (1 line) is enough since the remaining conditions already ensure we only match Option<Enum> patterns:

• anyOf with exactly 2 elements
• First has enum without nullable
• Second is { "enum": [null], "nullable": true }

The module restructuring could be a good follow-up, but I'd prefer to keep the bug fix minimal.

[clux]

quick first pass. i think this is a much better size of PR. thank you for your patience.

appreciate you're not moving to too much code around

hopes that the other transformers and their tests will be moved into individual modules [..] but decided to wait

yeah, let's wait until after this PR for that please.

[clux]

if possible. when checking for length, you should ideally destructure like in the original impl

        let [first, second] = arr.as_slice() else {
            return;
        };

to avoid a naked [0] and [1] later (which is only safe if you look 10 lines up)

but i don't know if this makes sense with the impl below (follow-up comment)

[clux]

this map needs a comment for what it's doing

[clux]

this comment is a moved variant of the old implementation, but it's only talking about the old "specific case" even though the implementation has been completely changed. is this right?