Document that <- is a single token

[mattheww]

<- was the "move operator", removed in Rust 0.5, but is still lexed as a single token.

So I think it should be listed in the Punctuation table (like the tilde token is).

Originally reported as ferrocene/specification#452

Fixes #583

[ehuss]

Thanks, yea, I noted this in #583.

Since there was an attempt to remove this that was then later backed out, I'm assuming the language team agrees that the emplacement operator has to stay permanently. However, I'm not 100% sure based on rust-lang/rust#50832 (comment).

Since I'm not completely confident that the language team doesn't want to try to remove this in the future, I'm going to nominate this for their attention for approval. I think this PR probably should be merged. I wouldn't be surprised if there were more macros using it since the last time it was attempted to be removed.

[mattheww]

It looks like the main question in rust-lang/rust#50832 is whether to remove this feature from the parser.

Perhaps the question of whether it will be removed from the lexer is easier.

In particular, if the answer to that one is "perhaps, but if so it would happen in a future edition", then I think there'd be no reason to delay documenting the current behaviour any further.

[scottmcm]

@rfcbot merge

I agree that we should reflect what's actually happening in the reference.

(And I think that it's not worth trying to change the lexing rules for this at this point.)

[rfcbot]

Team member @scottmcm has proposed to merge this. The next step is review by the rest of the tagged team members:

• @joshtriplett
• @nikomatsakis
• @pnkfelix
• @scottmcm
• @tmandry

Concerns:

• can we still remove the token resolved by Document that <- is a single token #1424 (comment)

Once a majority of reviewers approve (and at most 2 approvals are outstanding), this will enter its final comment period. If you spot a major issue that hasn't been raised at any point in this process, please speak up!

cc @rust-lang/lang-advisors: FCP proposed for lang, please feel free to register concerns.
See this document for info about what commands tagged team members can give me.

[tmandry]

Since there was an attempt to remove this that was then later backed out, I'm assuming the language team agrees that the emplacement operator has to stay permanently. However, I'm not 100% sure based on rust-lang/rust#50832 (comment).

It isn't clear to me why the token removal was backed out; the crater run above rust-lang/rust#60803 (comment) didn't find any actual regressions and the comment didn't leave a rationale.

It seems like it would have been possible to remove the token then, but now it might be harder. It seems worth running crater again to check, if only to make x<-1 parse correctly. Though if someone knows of crates that will definitely break or wants to argue for why it's a bad idea, we can skip it and I'll resolve the concern.

@rfcbot concern can we still remove the token

Otherwise, I'm not opposed to documenting the current behavior.

@rfcbot reviewed

[ehuss]

It sounds like from rust-lang/rust#60803 that Centril and petrochenkov wanted to keep the <- token for "future-proofing" and "keeping options open". Does the lang team want to close that door?

Perhaps it could be removed, and then added again later in an edition (though I don't know how feasible that is)?

[scottmcm]

Hmm, it looks like -> => <- and <= are all single-token today: https://play.rust-lang.org/?version=stable&mode=debug&edition=2021&gist=c068b8a62888b15c289e124ff39aebef

macro_rules! foo {
    ($x:tt) => {};
}

fn main() {
    foo!(->);
    foo!(=>);
    foo!(<-);
    foo!(<=);
}

So I think I'm now being pushed further to "let's just leave it as is", so tt-muncher macros can all keep using these as arrows if they want, since there's very little cost to keeping it.

[tmandry]

Discussed in t-lang and we agreed to unblock documenting this for now. @compiler-errors also wants to do a crater run so we can consider removing it, but that doesn't need to block this PR.

@rfcbot resolve can we still remove the token

[compiler-errors]

The crater run: rust-lang/rust#118920 (comment)

[traviscross]

@rustbot labels -I-lang-nominated

As mentioned above, we discussed this today and the consensus was that we should separate the question of documenting the current behavior from maybe changing that behavior. This can proceed as soon as enough people review and approve.

Once the crater run is analyzed, if someone wants to make a case for changing this behavior, please discuss the pros/cons and nominate for T-lang over in rust-lang/rust#118920 or in a new issue.

[ehuss]

The crater analysis was done in rust-lang/rust#118920, and @compiler-errors decided that there was enough breakage to not make it worth it.

I'm going to consider the comment in #1424 (comment) to imply that the team as a whole has agreed to this even though the rfcbot checkboxes weren't checked. I don't know who was in attendance, but I assume it was more than just two.