Feature: Implement the Body trait for more native data types to improve ergonomics

[akneni]

Currently, the 'String' is the only native data type that implements hyper::body::Body. If we're passing other data types as a body, we need to add the http_body_util crate and add a little bit of boiler plate when building our request.

use http_body_util::Full;

let some_bytes = vec![0_u8, 1, 2];
let mut req = Request::builder()
    .method("POST")
    .body(Full::<Bytes>::from(some_bytes))
    .unwrap();

Note: While the request builder itself may not require the body to implement the Body trait, both the sender and connection returned by http1::handshake require the request's body to implement it.

With the body trait implemented for Vec<u8> (just one example that probably should implement body) we could leave out the http_body_util crate. Building a request would also become more intuitive and ergonomic.

    let some_bytes = vec![0_u8, 1, 2];
    let mut req = Request::builder()
        .method("POST")
        .body(some_bytes)
        .unwrap();

[zaira-bibi]

Hi! I would like to try my hand on this issue.

[seanmonstar]

I've been personally torn on this. I appreciate the explicitness of having Full, and I kind of regret that we stabilized with an implementation on String.

So, part of me resists adding it to any other standard buffer types.

But I also understand that one exists, so should we just swallow the pill and provide the rest? Maybe. I don't love it. But perhaps I'm just wrong.

[akneni]

I've launched a poll here #3747 . I say we give it a little bit and see what others think.

[akneni]

It seems like 80% of people (granted, only out of 5 responses) want the body trait implemented on additional types. I can get started working on this if that's good with you @seanmonstar

[dswij]

I've been personally torn on this. I appreciate the explicitness of having Full, and I kind of regret that we stabilized with an implementation on String.

So, part of me resists adding it to any other standard buffer types.

But I also understand that one exists, so should we just swallow the pill and provide the rest? Maybe. I don't love it. But perhaps I'm just wrong.

In hindsight, it makes sense to at least have an implementation for bytes if we have one for String.

[zaira-bibi]

Are there any plans to have an implementation for more native data types? If yes, then I would love to work on this.

[akneni]

I've already implemented this trait for Vec<u8>, but no other native types. However, it has yet to be merged, so I think the maintainers are still hesitant to extend this past the string type.