RFC: Remove inner attributes and inner doc-comments syntax

[liigo]

The problem

There is a big pain to distinguish inner- and outer- attributes, the same as doc-comments, for a long time. Two different syntax, make it no clarity and no consistency. Especially, the semicolon in inner-attr is most ambiguous (for newbies). They are many comments in #2569 about this.

The solution

We remove the inner-attributes and inner-doc-comments, and we always use outer ones everywhere. In the past, there was no place to write crate-scope outer-attributes and outer-doc-comments. But now, crate is already a keyword, we can reuse it to define crates (#12107), and place attributes and doc-comments on top of it:

src/libstd/lib.rs:

/// The std library's documents, and attributes, go here
#[crate_type="rlib"]
#[crate_id="std"]
pub crate std;

As a result, inner attributes and doc-comments are not required any more, we can remove them entirely. Perhaps we should keep the //! comment syntax, if most people like it (I do).

[ghost]

Why do people dislike them so much? I love being able to write

fn foo() { #[inline];
    // ...
}

fn bar() { #[test];
    // ...
}

[liigo]

Because of ugly
2014年2月17日 下午7:09于 "György Andrasek" notifications@github.com写道：

Why do people dislike them so much? I love being able to write

fn foo() { #[inline];
// ...}
fn bar() { #[test];
// ...}

—
Reply to this email directly or view it on GitHubhttps://github.com//issues/12341#issuecomment-35247574
.

[lilyball]

I strongly disagree with this proposal. Besides requiring pub crate foo; to define crate-level attributes (which I think is pretty bad), this would also prevent modules from defining their own documentation/attributes and instead requiring them to be specified in their parent module's file.

I also prefer the ability to mark up some attributes inside a function instead of outside, if I consider them implementation details.

[emberian]

Closing. If you would like to pursue this further, please create an RFC using the new RFC process