macros: it is possible to downcast a `tt` to most anything using nested macros

[apoelstra]

The following code compiles:

#![feature(macro_rules)]

macro_rules! print_doccomment(
  (
    $doc:tt
    fn $name:ident()
  ) => (
    macro_rules! dummy( () => (
      $doc
      fn $name() {
        println!("The doc for this function is `{}`", stringify!($doc));
      }
    ))
    dummy!()
  )
)

print_doccomment!(
  /// This is a doc comment
  fn function_to_doc()
)

fn main() {
  function_to_doc();
}

And prints "/// This is a doc comment". However, the token /// This is a doc comment is not an item, which is the only thing that should parse for the first $doc when dummy! is expanded (according to @chris-morgan on IRC).

Note that I am using (a variation on) this in some code to transmit the doc comments on RPC-handling functions in response to the RPC "help". It would be great if there were a sanctioned way to parse or construct doccomments in macros.

[huonw]

I'm not sure I understand

However, the token /// This is a doc comment is not an item, which is the only thing that should parse for the first $doc when dummy! is expanded

Could you clarify? (The doc comment is an attribute that ends up attached to the fn.)

[bltavares]

Triaging: I was not able to compile on nightly. The code needed some ; added, but now it fails with an error on parsing.

<anon>:19:3: 19:28 error: no rules expected the token `[`
<anon>:19   /// This is a doc comment
            ^~~~~~~~~~~~~~~~~~~~~~~~~

#![feature(macro_rules)]

macro_rules! print_doccomment(
  (
    $doc:tt
    fn $name:ident()
  ) => (
    macro_rules! dummy( () => (
      $doc
      fn $name() {
        println!("The doc for this function is `{}`", stringify!($doc));
      }
    ))
    dummy!()
  )
);

print_doccomment!(
  /// This is a doc comment
  fn function_to_doc()
);

fn main() {
  function_to_doc();
}

[Mark-Simulacrum]

The original and updated code no longer compile, and I suspect it should be closed, since the original issue is, and I cannot reproduce.

@jseyfried Can you confirm?

[Mark-Simulacrum]

Closing. Please reopen if that's not the correct decision here.

[jseyfried]

Yeah, this should be closed -- we can now downcast tts anywhere.