Add support for counting syntactic matches in a macro definition

[ghost]

This is a proposed solution for #12782.

[huonw]

It would be good to have some tests for invalid and other edge cases too, e.g. all the cases in:

macro_rules! test {
     ($($($e: expr),*);*, $x: ident) => {
         (#($e),
          #($($e)*), 
          #($($($e)*)*),
          #($($e $e)*),
          #($($x $e)*),
          #($x),
          #($x $x),
          #($e $x),
          #(1))
          // etc.
     }
}

None of those should crash the compiler, and, preferably, they would do something sensible.

[ghost]

@huonw Actually, I should have made it clear that with this change, only the occurrences of an individual variable can be counted, not a whole pattern. So most of your examples will not parse.

I did add one more test for nested sequences which, as it turns out, didn't work properly.

[ghost]

@huonw Any thoughts on this one?

[huonw]

Oh, sorry (there's no notifications for when a PR is updated, so feel free to comment when you do so).

So most of your examples will not parse

If they are meant to emit compiler errors, you should add some compile-fail tests (explanation) to make sure that the compiler is doing the correct thing.

[emberian]

cc @paulstansifer

[alexcrichton]

We decided in today's meeting that while amenable, we'd like to see an RFC for this change to garner some discussion around it. It's unclear how well this interacts with the rest of the macro system, and we feel that an RFC would help flesh it out some more.

You can read up on the rfc process over at https://github.com/rust-lang/rfcs with RFC #1.

Thanks again for this PR, I'd love to see this get implemented!