Now that underscores are accepted in unicode escapes in Rust 1.22, should the UNICODE_ESCAPE production include underscores? Maybe something like this?

UNICODE_ESCAPE :
    \u{ HEX_DIGIT _* }
  | \u{ HEX_DIGIT _* HEX_DIGIT _* }
  | \u{ HEX_DIGIT _* HEX_DIGIT _* HEX_DIGIT _* }
  | \u{ HEX_DIGIT _* HEX_DIGIT _* HEX_DIGIT _* HEX_DIGIT _* }
  | \u{ HEX_DIGIT _* HEX_DIGIT _* HEX_DIGIT _* HEX_DIGIT _* HEX_DIGIT _* }
  | \u{ HEX_DIGIT _* HEX_DIGIT _* HEX_DIGIT _* HEX_DIGIT _* HEX_DIGIT _* HEX_DIGIT _* }

should the UNICODE_ESCAPE production include underscores? Maybe something like this?

Cool! I wasn't aware of that :). Yes, that would be the way I would include it in the grammar.

So... I am against including this right now. The problem in including this change is that I'm testing the grammar on stable to avoid having such a fast moving target and to avoid including in it things that maybe will took too long or never reach stable at all (since in my understanding lots of things on nightly are experimental).

Also, correct me if I am wrong: the Rust 1.22 you talk about is actually nightly, right? It doesn't mean this will be released on Rust 1.22, right? The underscore in Unicode escapes isn't available even in beta yet (Rust 1.21.beta3).

Rust 1.22 you talk about is actually nightly, right?

Nightly, yes.

BTW, do you want some help on this? I'd be willing to help if you could point me to things to add/fix. I also might have some questions about the grammar you have so far, and am curious how you're testing it.

The underscores in Unicode escapes is insta-stable (there's no feature gate) and the reference rides the release trains, so it should be included in this grammar. It can be done in a separate PR if that would help.

@Havvy

LGTM, minus the underscores from other comments

I've added them. Shouldn't they already be on beta, then? Or is it too soon?

@ehuss

BTW, do you want some help on this? I'd be willing to help if you could point me to things to add/fix.

Yeah, that would be great. But I have too many things half done... I will need to take a look at what I have and commit it.

I also might have some questions about the grammar you have so far

I'm on IRC sometimes, on #rust-docs .

and am curious how you're testing it.

I'm reading the rustc's parser/lexer, making an Antlr4 grammar out of it, and making some test cases and testing them against my grammar and rustc.

Now it is a little too long. I didn't want to use {n, m} notation but maybe

\u{ ( HEX_DIGIT _* )^{1,6} }

would be a better way to do this? Or even

\u{ ( HEX_DIGIT _* )^1..6 }

Yeah. It's literally one character too long when I render it, and spilling onto the next line in a confusing way. 😕 I don't care whether it's ^{n,m} or ^n..m, or even ^{1 to 6 repetitions} just as long as it conveys what is going on. It might be useful to describe the notation used by the grammar in lexical-structure.html including any interesting things you use.

💟 Thanks.