`async` and other Rust 2018+ keywords not recognized as so in Rustdoc highlighting

[ThePuzzlemaker]

I tried this code:

/// ```rust ignore
/// fn foo(bar: &dyn Baz);
/// async fn quux() {
///     qaz.frobnify().await;
/// }
/// fn bar() {
///     let mut a = 0;
///     abstract blah; /// is a reserved keyword
/// }
/// ```
fn foo() {}

I expected to see this happen: async and other Rust 2018+ keywords are highlighted as my project is set to edition = "2018" in its Cargo.toml

Instead, this happened: They were not highlighted, but regular keywords (including reserved ones) were.

I believe that the reason why this is happening is this snippet in librustdoc:

rust/src/librustdoc/html/highlight.rs

Lines 292 to 305 in eb4fc71

	TokenKind::Ident => match text {
	"ref" | "mut" => Class::RefKeyWord,
	"self" | "Self" => Class::Self_,
	"false" | "true" => Class::Bool,
	"Option" | "Result" => Class::PreludeTy,
	"Some" | "None" | "Ok" | "Err" => Class::PreludeVal,
	// Keywords are also included in the identifier set.
	_ if Ident::from_str(text).is_reserved() => Class::KeyWord,
	_ if self.in_macro_nonterminal => {
	self.in_macro_nonterminal = false;
	Class::MacroNonTerminal
	}
	_ => Class::Ident,
	},

In this case, Ident::from_str will only create a dummy span (which isn't set to 2018 edition), so is_reserved does not include 2018+ keywords.

I would be willing to try to work on a PR to fix this issue if it is deemed worthy.

Meta

Tested

rustc --version --verbose:
stable:

rustc 1.48.0 (7eac88abb 2020-11-16)
binary: rustc
commit-hash: 7eac88abb2e57e752f3302f02be5f3ce3d7adfb4
commit-date: 2020-11-16
host: x86_64-unknown-linux-gnu
release: 1.48.0
LLVM version: 11.0

nightly:

rustc 1.50.0-nightly (7efc097c4 2020-12-12)
binary: rustc
commit-hash: 7efc097c4fe6e97f54a44cee91c56189e9ddb41c
commit-date: 2020-12-12
host: x86_64-unknown-linux-gnu
release: 1.50.0-nightly

Backtrace is not applicable.

[ThePuzzlemaker]

@rustbot label +T-rustdoc +A-rustdoc-ui +A-async-await
Hope these labels are accurate and helpful.

[tmandry]

@ThePuzzlemaker I think you have the right idea, though someone from the Rustdoc team might also be able to help with the details (cc @jyn514).

A PR would be welcome! Feel free to claim the issue with @rustbot claim. If you have questions, I'm also available to help here or on Zulip or DIscord, but my availability may be spotty over the next week.

[ThePuzzlemaker]

@rustbot claim

[jyn514]

Yup, that looks right, good catch! The only tricky thing will be passing in a span in to Classifier::new, but I think all the callsites have access to at least a Session, and you can get an edition from there.

[ThePuzzlemaker]

@rustbot release-assignment
I'm going to release assignment as I'm still not quite sure how this could be achieved even after looking through a bit of code.

[jyn514]

@ThePuzzlemaker where did you get stuck? I might be able to help.

[ThePuzzlemaker]

@jyn514 I couldn't find a way to get access to a Session or how I even get a Span from the lexer. I guess I'll reclaim though.
@rustbot claim