fix: Fix return type of async closures. #12958
Conversation
crates/hir-ty/src/infer/expr.rs
Outdated
| // Use the first type parameter as the output type of future. | ||
| // existential type AsyncBlockImplTrait<InnerType>: Future<Output = InnerType> | ||
| let impl_trait_id = crate::ImplTraitId::AsyncBlockTypeImplTrait(self.owner, *body); | ||
| let opaque_ty_id = self.db.intern_impl_trait_id(impl_trait_id).into(); | ||
| dbg!(&ret_ty); | ||
| TyKind::OpaqueType(opaque_ty_id, Substitution::from1(Interner, ret_ty)) | ||
| .intern(Interner) |
There was a problem hiding this comment.
This was mostly copied from the Expr::Async arm on line 152 above.
There was a problem hiding this comment.
This uses AsyncBlockTypeImplTrait but isn't an async block.
Should a new variant be added to ImplTraitId? Or maybe AsyncBlockTypeImplTrait should be renamed?
|
Actually, this doesn't yet fix the issue, it gives (I also forgot to tidy, oops) |
zachs18
force-pushed
the
async_closure
branch
2 times, most recently
from
20d3b5a
to
2a6bdd1
Compare
| let f = async move |x: i32| x + 42; | ||
| f; | ||
| // ^ |i32| -> impl Future<Output = i32> |
There was a problem hiding this comment.
This probably shouldn't need x: i32, but otherwise it deduces the type as |i32| -> impl Future<Output = {unknown}>. It's not an issue with this PR though, and it also happens with non-async closures.
|
It now gives correct hints for the example in #12957.
(The |
|
☔ The latest upstream changes (presumably #13165) made this pull request unmergeable. Please resolve the merge conflicts. |
zachs18
force-pushed
the
async_closure
branch
2 times, most recently
from
2afae76
to
f9e8c6e
Compare
|
☔ The latest upstream changes (presumably #13209) made this pull request unmergeable. Please resolve the merge conflicts. |
Veykril
added
the
S-waiting-on-review
|
Can someone knowing the type system side of r-a have a look at this? |
|
There landed some changes revolving the closure inference code (we have landed support for generators), can you update this PR to make use of the same machinery? That will probably tell if your current approach works out fine as well. |
Veykril
added
S-waiting-on-author
|
Yes, I should have time this weekend to look into that. |
zachs18
force-pushed
the
async_closure
branch
from
f9e8c6e
to
49897ee
Compare
|
I rebased, but I didn't change to use the new machinery yet. I should be able to look into that sometime this week. |
|
☔ The latest upstream changes (presumably #14222) made this pull request unmergeable. Please resolve the merge conflicts. |
|
I went ahead and moved this to the new machinery, I hope you don't mind :) |
|
@bors r+ |
|
☀️ Test successful - checks-actions |
May fix #12957