feat: support declaring DTOField via Annotated

[peterschutt]

E.g.:

class A(Struct):
    a: Annotated[int, dto_field("read_only")]

For #2351

Pull Request Checklist

• New code has 100% test coverage
• (If applicable) The prose documentation has been updated to reflect the changes introduced by this PR
• (If applicable) The reference documentation has been updated to reflect the changes introduced by this PR

Description

Close Issue(s)

[provinzkraut]

@peterschutt this seems to be ready?

[peterschutt]

@peterschutt this seems to be ready?

Yeh close - no tests. I have some related changes to PR, however I'm AFK for the weekend.

[peterschutt]

If an argument to Annotated is not hashable, e.g., Annotated[int, {"set"}], then msgspec.inspect.type_info() throws a TypeError.

I created jcrist/msgspec#565 for that, however, for the tests to pass we'd either need to wait for that to make a release, or make DTOField hashable. I've chosen to do the latter. Any objections @litestar-org/maintainers?

[JacobCoffee]

Is that unsafe_hash needed for msgpsec?

[peterschutt]

Is that unsafe_hash needed for msgpsec?

Yep. Msgspec passes the type into dict.get() which calls _AnnotatedAlias.__hash__() which in turn passes a tuple of the instance and its metadata to hash(). This means that for Annotated to be hashable, all of its metadata must be hashable, and DTOField isn't.

I ended up going with unsafe_hash=True as I figure it would be less imposing than making the type eq=True, frozen=True. Having said that though, these things are pretty much designed to be instantiated in the global namespace (i.e., at model definition time), and it would be weird to modify them after the fact (although we do modify them to mark read-only et al. based on the DTOConfig configuration). So, I'd prefer to make them truly immutable and safe hash, but not sure if that would be too backward incompatible.

The other option is to block this until msgspec cuts their next release and not modify the hashability of the types at all.

[peterschutt]

The other option is to block this until msgspec cuts their next release and not modify the hashability of the types at all.

FYI: waiting for this.

[sonarcloud]

SonarCloud Quality Gate failed.   

0 Bugs
0 Vulnerabilities
0 Security Hotspots
0 Code Smells

91.3% Coverage
0.0% Duplication

Catch issues before they fail your Quality Gate with our IDE extension SonarLint

[github-actions]

Documentation preview will be available shortly at https://litestar-org.github.io/litestar-docs-preview/2367

[provinzkraut]

@peterschutt Anything blocking this aside from time constraints?

[peterschutt]

@peterschutt Anything blocking this aside from time constraints?

Nothing - except that I keep getting distracted with work closer to the top of the pile.

[peterschutt]

Superseded by #3289 3289