Why there is an implementation of IntoResponse for ()

[miaomiaowu0428]

Summary

Just as I mentioned in the title, I found this in axum:

impl IntoResponse for () {
    fn into_response(self) -> Response {
        Body::empty().into_response()
    }
}

What if someone simply want's to return "success" like this

 fn some_func() -> impl IntoResponse {
    "success".into_response()
}

but make a mistake like this:

 fn some_func() -> impl IntoResponse {
    "success".into_response();
}

then he/she will found this function always response a empty body with code 200
the mistake is easy to happen too inconspicuous to be spotted at the first time

If someone want's to make a response with an empty body, he/she shoule use StatusCode.into_response(), right?

axum version

0.7.5

[mladedav]

I guess this is a quality of life for simple handlers like async fn handle() { println!("I'm handling things!"); }

Some tests use it for example like this: let two = Router::new().route("/baz", get(|| async {}));

But I have to agree this might be something worth at least considering deprecating. In the meanwhile I'd suggest using a concrete return type rather than RPIT. This also shouldn't be an issue if you return from multiple places an make a mistake just once like

 fn some_func() -> impl IntoResponse { // ERROR: mismatched types expected struct `http::Response<body::Body>` found unit type `()`
    if false {
        return "failure".into_response();
    }
    "success".into_response();
}

As a side note, since these types already implement IntoResponse, there's no need to call into_response on them explicitly. Just return "success", (), StatusCode::OK or whatherever else. You should have to call this explicitly only if you want to return different types from different places, effectively returning resolved Responses instead.

[jplatte]

Maybe we should add #[must_use] to fn into_response.

[jplatte]

Once #2846 is merged and released, your example and similar code will generate a warning:

[miaomiaowu0428]

nice!