rustdoc: doc(hidden) also hides re-exports

[ghost]

Code:

#[macro_use]
pub mod my_module {
    /// This is a simple macro.
    #[macro_export]
    #[doc(hidden)]
    macro_rules! foo {
        () => ();
    }

    /// This doc comment doesn't show up in rustdoc (there's just an ugly re-export).
    pub use crate::foo as bar;

    // And this re-export doesn't show up in rustdoc at all!
    #[doc(inline)]
    pub use crate::foo as baz;
}

Generated docs:

So what I want to do is create a macro inside my_module. I don't want it to show up in crate root - I want it only in this module.

Both of the re-exports above work just fine. However, in the generated docs, one of them has no docs and the other doesn't show up at all.

I'm surprised that #[doc(hidden)] is "recursive" - shouldn't it only hide foo and not hide baz? Is this a bug?

Versions:

$ rustc --version
rustc 1.35.0-nightly (94fd04589 2019-03-21)
$ cargo --version
cargo 1.35.0-nightly (0e35bd8af 2019-03-13)
$ rustdoc --version
rustdoc 1.35.0-nightly (94fd04589 2019-03-21)

[weiznich]

Is there any news on this?
We would like to use something like that for diesel to.

[jonhoo]

This doesn't appear to be just an issue with macros either:

/// Hello world
#[doc(hidden)]
pub struct Foo;

#[doc(inline)]
pub use Foo as Bar;

Produces entirely empty rustdoc output!

[nazar-pc]

I have a similar issue with modules:

pub mod foo {
    #[doc(hidden)]
    mod bar {
        pub struct Baz;
    }
}

#[doc(inline)]
pub use foo::bar;

Essentially I want to present nested module structure (that is used to limit visibility within crate) as non-nested in docs by leveraging re-exports.

I'm thinking of #[doc(visible)] that would override #[doc(hidden)] of the definition if applied to pub use:

pub mod foo {
    #[doc(hidden)]
    mod bar {
        pub struct Baz;
    }
}

#[doc(inline, visible)]
pub use foo::bar;

Without this when inlining is used it creates confusion since the same module is both visible at the root of the crate and nested within submodule. This also confuses IntelliJ Rust, it shows 2 suggestions for imports all the time and could use this to hide one of them.

[SergioBenitez]

This is causing quite a big issue for Rocket's docs. We'd really like to export macro_rules! macros only from specific modules, but this bug makes it impossible without also exporting them at the root.

Specifically, this:

pub mod foo {
    /// `macro_rules!`
    #[macro_export]
    macro_rules! m { () => () }

    /// `pub use m;`
    #[doc(inline)]
    pub use m;
}

Results in:

index:

modules:
foo

macros:
m `macro_rules!`

foo:

macros:
m `macro_rules!`

Which is expected. Now, by adding #[doc(hidden)] to m's declaration/definition...

    #[macro_export]
+   #[doc(hidden)]
    macro_rules! m { () => () }

...all instances of m in the docs disappear, even though foo:m is still reachable and not hidden!

[SergioBenitez]

Here's a macro that (sort of) patches around this issue:

#[macro_use]
macro_rules! export {
    (
        $(#[$m:meta])*
        macro_rules! $name:ident {
            $(($($token:tt)*) => ($($expansion:tt)*));* $(;)?
        }
    ) => (
        $(#[$m])*
        #[doc(hidden)]
        #[cfg(not(nightly))]
        #[macro_export]
        macro_rules! $name {
            $(
                ($($token)*) => ($($expansion)*)
            );*
        }

        $(#[$m])*
        #[cfg(not(nightly))]
        pub use $name;

        $(#[$m])*
        #[cfg(nightly)]
        pub macro $name {
            $(
                ($($token)*) => ($($expansion)*)
            ),*
        }
    )
}

Simply wrap a macro_rules! declaration with export!:

export! {
    macro_rules! my_macro { .. }
}

One must be careful not to use decl_macro features in expansion arms. As long as this issue exists, the macro will appear as pub use macro; in the right place without a docstring. Should this issue be fixed, nightly and stable docs will match.

[SpriteOvO]

It looks like this issue has been inactive for a while, but the problem is still there.

I would like to know if there are more comments from members about #59368 (comment)?

@danielhenrymantilla said in this reply he had a discussion with @jyn514 and @GuillaumeGomez about this, but I didn't find it. @jyn514 replied later that it looked fine to him.

[jyn514]

I'm no longer a member of the rustdoc team, please don't ping me.

[GuillaumeGomez]

Technical information for everyone following this issue: I'm working on it (well, you can infer that from my PRs I suppose). One of the last blockers is actually #[doc(hidden)]: currently we merge both the pub use's and the item's attributes. Meaning that the #[doc(hidden)] is inherited. The problem is that it's currently not possible to filter the sub-attributes of an ast::Attribute (for example, in #[doc(hidden, cfg(feature = "foo"))], we want to keep cfg(feature = "foo")).

So hopefully, once I figure out this part, I should be able to fix all the remaining issues at once.

[GuillaumeGomez]

So now only one issue remains:

#[macro_export]
#[doc(hidden)]
macro_rules! foo {
    () => {};
}

pub use crate::foo as Macro; // this one isn't displayed, as expected
#[doc(inline)]
pub use crate::foo as Macro2; // this one isn't displayed but should be!

[danielhenrymantilla]

FWIW, "back in the day" I did have a branch with an implementation tackling your Macro2 situation, @GuillaumeGomez, if you are interested: danielhenrymantilla@c68d225

[GuillaumeGomez]

I have a vague idea on how to fix it but I'm definitely interested into seeing how you did it. You sure you don't want to send a PR btw?