async T and gen T types

text/3628-async-gen-types.md

@@ -0,0 +1,154 @@
+- Feature Name: `async_gen_types`
+- Start Date: 2024-05-06
+- RFC PR: [rust-lang/rfcs#3628](https://github.com/rust-lang/rfcs/pull/3628)
+- Rust Issue: [rust-lang/rust#0000](https://github.com/rust-lang/rust/issues/0000)
+
+# Summary
+[summary]: #summary
+
+Allow the syntax `async T` and `gen T` as types, equivalent to
+`impl Future<Output = T>` and `impl Iterator<Item = T>` respectively. Accept
+them anywhere `impl Trait` can appear.
+
+# Motivation
+[motivation]: #motivation
+
+Users working with asynchronous code may encounter `impl Future<Output = T>`
+types.  Users working with iterators may encounter `impl Iterator<Item = T>`
+types.
+
+These types are long and cumbersome to work with. They may be the first time
+a user will encounter an associated type, and they add verbosity that
+obfuscates the `Output`/`Item` types that people care more about. In
+particular, a function that combines multiple futures or iterators with other
+types requires reading past a lot of syntactic overhead.
+
+Users do not encounter these types when consuming iterators with loops or
+combinators (or in the future producing them with `gen` blocks), or when
+producing or consuming futures using async/await syntax.
+
+The syntax proposed by this RFC provides the same benefits that the current
+`async fn` syntax does (highlighting the future output type), but usable in any
+type rather than only in function return values.
+
+# Explanation
+[explanation]: #explanation
+
+In any context where you can write an `impl Trait` type, you can write
+`async T`, which desugars to `impl Future<Output = T>`:
+
+```rust
+fn future_seq<T, U>(f1: async T, f2: async U) -> async (T, U) {
+    async {
+        (f1.await, f2.await)
+    }
+}
+```
+
+Similarly, in any context where you can write an `impl Trait` type, you can
+write `gen T`, which desugars to `impl Iterator<Item = T>`:
+
+```rust
+fn iter_seq<T>(g1: gen T, g2: gen T) -> gen T {
+    gen {
+        yield from g1;
+        yield from g2;
+    }
+}
+```
+
+These syntaxes work exactly as their desugarings suggest, and can appear
+anywhere their desugarings can appear.
+
+Compare these to the longhand versions of these two functions:
+
+```rust
+fn future_seq<T, U>(
+    f1: impl Future<Output = T>,
+    f2: impl Future<Output = U>,
+) -> impl Future<Output = (T, U)> {
+    async {
+        (f1.await, f2.await)
+    }
+}
+
+fn iter_seq<T>(
+    g1: impl Iterator<Item = T>,
+    g2: impl Iterator<Item = T>)
+-> impl Iterator<Item = T> {
+    gen {
+        yield from g1;
+        yield from g2;
+    }
+}
+```
+
+Notice how much longer these are, and how much more syntax the user needs to
+wade through to observe the types they care about.
+
+# Drawbacks
+[drawbacks]: #drawbacks
+
+This adds an additional case to Rust type syntax.
+
+# Rationale and alternatives
+[rationale-and-alternatives]: #rationale-and-alternatives
+
+We could introduce a mechanism to abbreviate `impl Future<Item = T>` as
+`impl Future<T>` or `impl Fut<T>` or similar. However, this still leaves much
+of the syntactic "weight" in place. In addition, this may confuse users by
+obfuscating the difference between associated types and generic parameters.
+
+# Prior art
+[prior-art]: #prior-art
+
+We have special syntaxes for arrays in types, `[T]` and `[T; N]`, which are
+evocative of the corresponding value syntax for arrays. Similarly, the syntax
+for tuple types `(A, B)` is evocative of the syntax for tuple values `(a, b)`.
+
+The use of `async fn` to hide the asynchronous type serves as a partial
+precedent for this: the case made at the time was that users cared about the
+output type of the future more than they cared about the `Future` trait. This
+RFC extends that benefit to any place a type can appear.
+
+Similarly, `async` blocks do not require specifying the Future trait, and
+neither do the proposed `gen` blocks.
+
+# Unresolved questions
+[unresolved-questions]: #unresolved-questions
+
+The `gen T` syntax can be added as unstable right away, but should not be
+stabilized until we stabilize the rest of `gen` support.
+
+Introducing `async T` as a type meaning `impl Future<Output = T>` would close
+off the use of `async T` as a syntax for "asynchronous versions" of existing
+types (e.g. `async File`).
+
+# Future possibilities
+[future-possibilities]: #future-possibilities
+
+Once we add `async gen` support, we can add the corresponding type
+`async gen T`, mapping to whatever type we use for async iterators.
+
+These syntaxes would work very well together with a syntax to abbreviate
+functions consisting of a single block.
+
+For example:
+
+```rust
+fn countup(limit: usize) -> gen usize
+gen {
+    for x in 0..limit {
+        yield i;
+    }
+}
+
+fn do_something_asynchronously() -> async ()
+async {
+    do_something().await;
+}
+```
+
+Together, these mechanisms would provide a general solution for what might
+otherwise motivate a `gen fn` feature. Using `gen T` as a type makes the return
+type simple enough to not need to hide the type.