[stdlib] Give the CollectionElementNew trait to UnsafeMaybeUninitalized

[gabrieldemarmiesse]

This is part of the PR #2965

@ConnorGray I think you're the right person for a review :)

[gabrieldemarmiesse]

@ConnorGray all done. I also think that it should be possible in a future PR to remove the requirements for the type that can be put inside an UnsafeMaybeUninitialized. And also to use "constrained" instead of abort

EDIT: Actually, I'll clean up more stuff in this PR.

[gabrieldemarmiesse]

Summary of the changes:

1. UnsafeMaybeUninitialized now takes AnyType as element. Added very specific traits to each method to allow this.
2. UnsafeMaybeUninitialized now has the CollectionElementNew trait, but none of the methods of this trait can actually be called because of ambiguity, and will fail at compile time.
3. Switch from abort to constrained for a faster fail. It's a nice QoL feature.
4. Removed inout from initialize_pointee_explicit_copy as we are not mutating the pointer but the value it points to. This is because, without this change, we can't do self.unsafe_ptr().initialize_pointee_explicit_copy(...) because it's mutating an rvalue, which is not allowed by the compiler (rightfully or not, I don't have an opinion on this).
5. Moved ValueDestructorRecorder in the file stdlib/test/test_utils/types.mojo where it can be used in other test files, it's a very useful struct.

@ConnorGray it's ready for review, for real this time :)

[ConnorGray]

!sync

[ConnorGray]

I think making this a compile time failure might prevent uses like InlineArray[UnsafeMaybeUninitialized[T]], because that type would instantiate a call to this copy constructor.

The goal would be to avoid dynamically doing a copy operation, but we need the compiler to accept code that "looks like" it could do a copy, because its up to the programmer to make sure no actual copy gets attempted.

[ConnorGray]

👍 , this makes sense.

[ConnorGray]

Hey Gabriel, thanks for your work on this, I think this all looks good—I like that change to make this work with AnyType.

Re: the constrainted[False] change, I think that will make InlineArray[UnsafeMaybeUninitialized[T]] fail to compile. Considering fixing InlineArray / InlineList was the motivation for this type, I think we might to stick with abort for the time being.

We could investigate switching this to constrained[False] maybe in a different PR. (Or maybe even removing the __init__(inout self, Self) and __moveinit__ constructors entirely?)

[ConnorGray]

Awesome, thank you for the changes Gabriel! This looks good, I'll move forward with merging :)

[ConnorGray]

!sync

[modularbot]

✅🟣 This contribution has been merged 🟣✅

Your pull request has been merged to the internal upstream Mojo sources. It will be reflected here in the Mojo repository on the nightly branch during the next Mojo nightly release, typically within the next 24-48 hours.

We use Copybara to merge external contributions, click here to learn more.

[modularbot]

Landed in 431a620! Thank you for your contribution 🎉