RFC: Attributes in function return type position

text/3201-return-type-attrs.md

@@ -0,0 +1,251 @@
+- Feature Name: `return_type_attrs`
+- Start Date: 2021-11-25
+- RFC PR: [rust-lang/rfcs#3201](https://github.com/rust-lang/rfcs/pull/3201)
+
+# Summary
+[summary]: #summary
+
+Allow attributes in function return type position. For example:
+
+```rust
+fn sum(x: u8, y: u8) -> #[some_attr] u8 {
+    x + y
+}
+```
+
+# Motivation
+[motivation]: #motivation
+
+Currently the whole function can be annotated with a macro attribute:
+
+```rust
+#[some_attr]
+fn example() { .. }
+```
+
+As well as individual function parameters (introduced in [RFC2565]):
+
+[RFC2565]: 2565-formal-function-parameter-attributes.md
+
+```rust
+fn example(#[attr] input: String, #[attr] x: u8) { .. }
+```
+
+However function return types currently cannot be annotated, which forces
+domain-specific languages (DSLs) to resort to function attributes, for example:
+
+```rust
+#[wasm_bindgen]
+impl RustLayoutEngine {
+    #[return_type = "MapNode[]"]
+    pub fn layout(
+        &self,
+        #[type = "MapNode[]"] nodes: Vec<JsValue>,
+        #[type = "MapEdge[]"] edges: Vec<JsValue>
+    ) -> Vec<JsValue> {
+        ..
+    }
+}
+```
+
+Return type attributes would allow the above example to be changed to:
+
+```rs
+#[wasm_bindgen]
+impl RustLayoutEngine {
+    pub fn layout(
+        &self,
+        #[type = "MapNode[]"] nodes: Vec<JsValue>,
+        #[type = "MapEdge[]"] edges: Vec<JsValue>
+    ) -> #[type = "MapNode[]"] Vec<JsValue> {
+        ..
+    }
+}
+```
+
+The resulting DSL is clearer to read ... in particular in functions with many
+parameters where the function attribute would be several lines apart from the
+return type (because `rustfmt` by default puts each parameter on its own line
+when the function signature gets too long).
+
+Since function parameters already can be annotated this can be regarded as the
+next logical step towards more expressive and intuitive DSLs.  The motivation
+for the introduction of parameter attributes outlined in [RFC2565] largely
+applies to return type attributes as well, since they would also be useful for
+property based testing, interoperability with other languages and optimization
+annotations.
+
+# Guide-level explanation
+[guide-level-explanation]: #guide-level-explanation
+
+Return types of `fn` definitions as well as closures may have attributes
+attached to them. Thereby, additional information may be provided.
+
+For the purposes of illustration, let's assume we have the attribute
+`#[apple]` available to us.
+
+## Basic examples
+
+The syntax for attaching attributes to return types is shown in the snippet below:
+
+```rust
+// Free functions:
+fn foo() -> #[apple] u32 { .. }
+
+impl Alpha { // In inherent implementations.
+    fn bar() -> #[apple] u8 { .. }
+
+    ..
+}
+
+impl Beta for Alpha { // Also works in trait implementations.
+    fn bar() -> #[apple] u8 { .. }
+
+    ..
+}
+
+fn foo() {
+    // Closures:
+    let bar = |x| -> #[apple] u8 { .. };
+}
+```
+
+## Trait definitions
+
+An `fn` definition doesn't need to have a body to permit return type attributes.
+Thus, in `trait` definitions, we may write:
+
+```rust
+trait Beta {
+    fn bar(&self) -> #[apple] u8;
+}
+```
+
+## `fn` types
+
+You can also use attributes in function pointer types.
+For example, you may write:
+
+```rust
+type Foo = fn() -> #[apple] u8;
+```
+
+## Unit return type
+
+When annotating the unit return type `()` must be specified explicitly. For
+example:
+
+```rust
+fn foo() -> #[apple] () {
+    ..
+}
+```
+
+Attempting the following:
+
+```rust
+fn foo() -> #[apple] {
+    ..
+}
+```
+
+will result in a compile error:
+
+```
+error: return type attributes require an explicit return type
+fn foo() -> #[apple] {
+                    ^ expected ()
+```
+
+## Built-in attributes
+
+Attributes attached to return types do not have an inherent meaning in
+the type system or in the language. Instead, the meaning is what your
+procedural macros, the tools you use, or what the compiler interprets certain
+specific attributes as.
+
+As for the built-in attributes and their semantics, we will, for the time being,
+only permit the following attributes on return types:
+
+- Lint check attributes, that is:
+  `#[allow(C)]`, `#[warn(C)]`, `#[deny(C)]`, `#[forbid(C)]`,
+  and tool lint attributes such as `#[allow(clippy::foobar)]`.
+
+All other built-in attributes will be rejected with a semantic check.
+For example, you may not write:
+
+```rust
+fn foo() -> #[inline] u32 { .. }
+```
+
+
+# Reference-level explanation
+[reference-level-explanation]: #reference-level-explanation
+
+TODO
+
+<!--
+This is the technical portion of the RFC. Explain the design in sufficient detail that:
+
+- Its interaction with other features is clear.
+- It is reasonably clear how the feature would be implemented.
+- Corner cases are dissected by example.
+
+The section should return to the examples given in the previous section, and
+explain more fully how the detailed proposal makes those examples work.
+-->
+
+# Drawbacks
+[drawbacks]: #drawbacks
+
+All drawbacks for attributes in any location also count for this proposal.
+
+Having attributes in many different places of the language complicates its
+grammar.
+
+# Rationale and alternatives
+[rationale-and-alternatives]: #rationale-and-alternatives
+
+> Why is this design the best in the space of possible designs?
+
+If function parameters can be annotated it is only natural that function return
+types can be annotated as well.
+
+> What other designs have been considered and what is the rationale for not choosing them?
+
+Using a function attribute to annotate the return type can result in the
+annotation being far apart from the type that it's annotating, as showcased in
+the [Motivation](#motivation) section.
+
+[RFC2602](https://github.com/rust-lang/rfcs/pull/2602) suggests permitting
+attributes to be attached to any types, so implementing that RFC would also
+permit return types to be annotated with attributes. The concern that has
+however been raised with that approach is that permitting attributes nearly
+everywhere would undesirably increase the cognitive load needed to read Rust
+code and thus harm the readability of Rust.  Currently attributes are
+restricted to specific positions. Allowing them anywhere (even in nested types)
+would pose a more radical change, whereas this RFC is more of a continuation of
+the status quo by just permitting attributes in one more specific position.
+
+> What is the impact of not doing this?
+
+DSLs cannot take advantage of the additional expressiveness.
+
+# Prior art
+[prior-art]: #prior-art
+
+Parameter attributes were introduced to Rust with [RFC2565].
+
+# Unresolved questions
+[unresolved-questions]: #unresolved-questions
+
+Are there more built-in attributes that should be permitted for return types?
+
+Is there precedent of other programming languages permitting return type
+annotations to be placed directly in front of the return type?
+
+# Future possibilities
+[future-possibilities]: #future-possibilities
+
+If `rustdoc` one day supports documentation comments on parameters, it could
+also support documentation comments on return types.