Rollup merge of rust-lang#126452 - compiler-errors:raw-lifetimes, r=spastorino

Implement raw lifetimes and labels (`'r#ident`)

This PR does two things:
1. Reserve lifetime prefixes, e.g. `'prefix#lt` in edition 2021.
2. Implements raw lifetimes, e.g. `'r#async` in edition 2021.

This PR additionally extends the `keyword_idents_2024` lint to also check lifetimes.

cc `@traviscross`
r? parser

Author: matthiaskrgr

compiler/rustc_ast/src/mut_visit.rs

@@ -752,7 +752,7 @@ fn visit_lazy_tts<T: MutVisitor>(vis: &mut T, lazy_tts: &mut Option<LazyAttrToke
 pub fn visit_token<T: MutVisitor>(vis: &mut T, t: &mut Token) {
     let Token { kind, span } = t;
     match kind {
-        token::Ident(name, _ /*raw*/) | token::Lifetime(name) => {
+        token::Ident(name, _is_raw) | token::Lifetime(name, _is_raw) => {
             let mut ident = Ident::new(*name, *span);
             vis.visit_ident(&mut ident);
             *name = ident.name;
@@ -762,7 +762,7 @@ pub fn visit_token<T: MutVisitor>(vis: &mut T, t: &mut Token) {
         token::NtIdent(ident, _is_raw) => {
             vis.visit_ident(ident);
         }
-        token::NtLifetime(ident) => {
+        token::NtLifetime(ident, _is_raw) => {
             vis.visit_ident(ident);
         }
         token::Interpolated(nt) => {

compiler/rustc_ast/src/token.rs

@@ -331,11 +331,11 @@ pub enum TokenKind {
     /// Do not forget about `NtLifetime` when you want to match on lifetime identifiers.
     /// It's recommended to use `Token::(lifetime,uninterpolate,uninterpolated_span)` to
     /// treat regular and interpolated lifetime identifiers in the same way.
-    Lifetime(Symbol),
+    Lifetime(Symbol, IdentIsRaw),
     /// This identifier (and its span) is the lifetime passed to the
     /// declarative macro. The span in the surrounding `Token` is the span of
     /// the `lifetime` metavariable in the macro's RHS.
-    NtLifetime(Ident),
+    NtLifetime(Ident, IdentIsRaw),
 
     /// An embedded AST node, as produced by a macro. This only exists for
     /// historical reasons. We'd like to get rid of it, for multiple reasons.
@@ -458,7 +458,7 @@ impl Token {
     /// if they keep spans or perform edition checks.
     pub fn uninterpolated_span(&self) -> Span {
         match self.kind {
-            NtIdent(ident, _) | NtLifetime(ident) => ident.span,
+            NtIdent(ident, _) | NtLifetime(ident, _) => ident.span,
             Interpolated(ref nt) => nt.use_span(),
             _ => self.span,
         }
@@ -661,7 +661,9 @@ impl Token {
     pub fn uninterpolate(&self) -> Cow<'_, Token> {
         match self.kind {
             NtIdent(ident, is_raw) => Cow::Owned(Token::new(Ident(ident.name, is_raw), ident.span)),
-            NtLifetime(ident) => Cow::Owned(Token::new(Lifetime(ident.name), ident.span)),
+            NtLifetime(ident, is_raw) => {
+                Cow::Owned(Token::new(Lifetime(ident.name, is_raw), ident.span))
+            }
             _ => Cow::Borrowed(self),
         }
     }
@@ -679,11 +681,11 @@ impl Token {
 
     /// Returns a lifetime identifier if this token is a lifetime.
     #[inline]
-    pub fn lifetime(&self) -> Option<Ident> {
+    pub fn lifetime(&self) -> Option<(Ident, IdentIsRaw)> {
         // We avoid using `Token::uninterpolate` here because it's slow.
         match self.kind {
-            Lifetime(name) => Some(Ident::new(name, self.span)),
-            NtLifetime(ident) => Some(ident),
+            Lifetime(name, is_raw) => Some((Ident::new(name, self.span), is_raw)),
+            NtLifetime(ident, is_raw) => Some((ident, is_raw)),
             _ => None,
         }
     }
@@ -865,7 +867,7 @@ impl Token {
                 _ => return None,
             },
             SingleQuote => match joint.kind {
-                Ident(name, IdentIsRaw::No) => Lifetime(Symbol::intern(&format!("'{name}"))),
+                Ident(name, is_raw) => Lifetime(Symbol::intern(&format!("'{name}")), is_raw),
                 _ => return None,
             },
 

compiler/rustc_ast/src/tokenstream.rs

@@ -482,11 +482,11 @@ impl TokenStream {
             token::NtIdent(ident, is_raw) => {
                 TokenTree::Token(Token::new(token::Ident(ident.name, is_raw), ident.span), spacing)
             }
-            token::NtLifetime(ident) => TokenTree::Delimited(
+            token::NtLifetime(ident, is_raw) => TokenTree::Delimited(
                 DelimSpan::from_single(token.span),
                 DelimSpacing::new(Spacing::JointHidden, spacing),
                 Delimiter::Invisible,
-                TokenStream::token_alone(token::Lifetime(ident.name), ident.span),
+                TokenStream::token_alone(token::Lifetime(ident.name, is_raw), ident.span),
             ),
             token::Interpolated(ref nt) => TokenTree::Delimited(
                 DelimSpan::from_single(token.span),

compiler/rustc_ast_pretty/src/pprust/state.rs

@@ -11,7 +11,9 @@ use std::borrow::Cow;
 use ast::TraitBoundModifiers;
 use rustc_ast::attr::AttrIdGenerator;
 use rustc_ast::ptr::P;
-use rustc_ast::token::{self, BinOpToken, CommentKind, Delimiter, Nonterminal, Token, TokenKind};
+use rustc_ast::token::{
+    self, BinOpToken, CommentKind, Delimiter, IdentIsRaw, Nonterminal, Token, TokenKind,
+};
 use rustc_ast::tokenstream::{Spacing, TokenStream, TokenTree};
 use rustc_ast::util::classify;
 use rustc_ast::util::comments::{Comment, CommentStyle};
@@ -947,8 +949,13 @@ pub trait PrintState<'a>: std::ops::Deref<Target = pp::Printer> + std::ops::Dere
             token::NtIdent(ident, is_raw) => {
                 IdentPrinter::for_ast_ident(ident, is_raw.into()).to_string().into()
             }
-            token::Lifetime(name) => name.to_string().into(),
-            token::NtLifetime(ident) => ident.name.to_string().into(),
+
+            token::Lifetime(name, IdentIsRaw::No)
+            | token::NtLifetime(Ident { name, .. }, IdentIsRaw::No) => name.to_string().into(),
+            token::Lifetime(name, IdentIsRaw::Yes)
+            | token::NtLifetime(Ident { name, .. }, IdentIsRaw::Yes) => {
+                format!("'r#{}", &name.as_str()[1..]).into()
+            }
 
             /* Other */
             token::DocComment(comment_kind, attr_style, data) => {

compiler/rustc_expand/src/mbe/macro_parser.rs

@@ -398,8 +398,10 @@ pub(crate) enum NamedMatch {
 fn token_name_eq(t1: &Token, t2: &Token) -> bool {
     if let (Some((ident1, is_raw1)), Some((ident2, is_raw2))) = (t1.ident(), t2.ident()) {
         ident1.name == ident2.name && is_raw1 == is_raw2
-    } else if let (Some(ident1), Some(ident2)) = (t1.lifetime(), t2.lifetime()) {
-        ident1.name == ident2.name
+    } else if let (Some((ident1, is_raw1)), Some((ident2, is_raw2))) =
+        (t1.lifetime(), t2.lifetime())
+    {
+        ident1.name == ident2.name && is_raw1 == is_raw2
     } else {
         t1.kind == t2.kind
     }

compiler/rustc_expand/src/mbe/transcribe.rs

@@ -283,9 +283,9 @@ pub(super) fn transcribe<'a>(
                             let kind = token::NtIdent(*ident, *is_raw);
                             TokenTree::token_alone(kind, sp)
                         }
-                        MatchedSingle(ParseNtResult::Lifetime(ident)) => {
+                        MatchedSingle(ParseNtResult::Lifetime(ident, is_raw)) => {
                             marker.visit_span(&mut sp);
-                            let kind = token::NtLifetime(*ident);
+                            let kind = token::NtLifetime(*ident, *is_raw);
                             TokenTree::token_alone(kind, sp)
                         }
                         MatchedSingle(ParseNtResult::Nt(nt)) => {

compiler/rustc_expand/src/proc_macro_server.rs

@@ -229,15 +229,16 @@ impl FromInternal<(TokenStream, &mut Rustc<'_, '_>)> for Vec<TokenTree<TokenStre
                     span: ident.span,
                 })),
 
-                Lifetime(name) => {
+                Lifetime(name, is_raw) => {
                     let ident = symbol::Ident::new(name, span).without_first_quote();
                     trees.extend([
                         TokenTree::Punct(Punct { ch: b'\'', joint: true, span }),
-                        TokenTree::Ident(Ident { sym: ident.name, is_raw: false, span }),
+                        TokenTree::Ident(Ident { sym: ident.name, is_raw: is_raw.into(), span }),
                     ]);
                 }
-                NtLifetime(ident) => {
-                    let stream = TokenStream::token_alone(token::Lifetime(ident.name), ident.span);
+                NtLifetime(ident, is_raw) => {
+                    let stream =
+                        TokenStream::token_alone(token::Lifetime(ident.name, is_raw), ident.span);
                     trees.push(TokenTree::Group(Group {
                         delimiter: pm::Delimiter::None,
                         stream: Some(stream),

compiler/rustc_lexer/src/lib.rs

@@ -91,6 +91,15 @@ pub enum TokenKind {
     /// tokens.
     UnknownPrefix,
 
+    /// An unknown prefix in a lifetime, like `'foo#`.
+    ///
+    /// Note that like above, only the `'` and prefix are included in the token
+    /// and not the separator.
+    UnknownPrefixLifetime,
+
+    /// `'r#lt`, which in edition < 2021 is split into several tokens: `'r # lt`.
+    RawLifetime,
+
     /// Similar to the above, but *always* an error on every edition. This is used
     /// for emoji identifier recovery, as those are not meant to be ever accepted.
     InvalidPrefix,
@@ -677,9 +686,17 @@ impl Cursor<'_> {
             return Literal { kind, suffix_start };
         }
 
+        if self.first() == 'r' && self.second() == '#' && is_id_start(self.third()) {
+            // Eat "r" and `#`, and identifier start characters.
+            self.bump();
+            self.bump();
+            self.bump();
+            self.eat_while(is_id_continue);
+            return RawLifetime;
+        }
+
         // Either a lifetime or a character literal with
         // length greater than 1.
-
         let starts_with_number = self.first().is_ascii_digit();
 
         // Skip the literal contents.
@@ -688,15 +705,17 @@ impl Cursor<'_> {
         self.bump();
         self.eat_while(is_id_continue);
 
-        // Check if after skipping literal contents we've met a closing
-        // single quote (which means that user attempted to create a
-        // string with single quotes).
-        if self.first() == '\'' {
-            self.bump();
-            let kind = Char { terminated: true };
-            Literal { kind, suffix_start: self.pos_within_token() }
-        } else {
-            Lifetime { starts_with_number }
+        match self.first() {
+            // Check if after skipping literal contents we've met a closing
+            // single quote (which means that user attempted to create a
+            // string with single quotes).
+            '\'' => {
+                self.bump();
+                let kind = Char { terminated: true };
+                Literal { kind, suffix_start: self.pos_within_token() }
+            }
+            '#' if !starts_with_number => UnknownPrefixLifetime,
+            _ => Lifetime { starts_with_number },
         }
     }
 

compiler/rustc_lint/messages.ftl

@@ -708,6 +708,10 @@ lint_range_endpoint_out_of_range = range endpoint is out of range for `{$ty}`
 lint_range_use_inclusive_range = use an inclusive range instead
 
 
+lint_raw_prefix = prefix `'r` is reserved
+    .label = reserved prefix
+    .suggestion = insert whitespace here to avoid this being parsed as a prefix in Rust 2021
+
 lint_reason_must_be_string_literal = reason must be a string literal
 
 lint_reason_must_come_last = reason in lint attribute must come last

compiler/rustc_lint/src/builtin.rs

@@ -1851,9 +1851,16 @@ impl KeywordIdents {
                 TokenTree::Token(token, _) => {
                     if let Some((ident, token::IdentIsRaw::No)) = token.ident() {
                         if !prev_dollar {
-                            self.check_ident_token(cx, UnderMacro(true), ident);
+                            self.check_ident_token(cx, UnderMacro(true), ident, "");
                         }
-                    } else if *token == TokenKind::Dollar {
+                    } else if let Some((ident, token::IdentIsRaw::No)) = token.lifetime() {
+                        self.check_ident_token(
+                            cx,
+                            UnderMacro(true),
+                            ident.without_first_quote(),
+                            "'",
+                        );
+                    } else if token.kind == TokenKind::Dollar {
                         prev_dollar = true;
                         continue;
                     }
@@ -1869,6 +1876,7 @@ impl KeywordIdents {
         cx: &EarlyContext<'_>,
         UnderMacro(under_macro): UnderMacro,
         ident: Ident,
+        prefix: &'static str,
     ) {
         let (lint, edition) = match ident.name {
             kw::Async | kw::Await | kw::Try => (KEYWORD_IDENTS_2018, Edition::Edition2018),
@@ -1902,7 +1910,7 @@ impl KeywordIdents {
         cx.emit_span_lint(
             lint,
             ident.span,
-            BuiltinKeywordIdents { kw: ident, next: edition, suggestion: ident.span },
+            BuiltinKeywordIdents { kw: ident, next: edition, suggestion: ident.span, prefix },
         );
     }
 }
@@ -1915,7 +1923,11 @@ impl EarlyLintPass for KeywordIdents {
         self.check_tokens(cx, &mac.args.tokens);
     }
     fn check_ident(&mut self, cx: &EarlyContext<'_>, ident: Ident) {
-        self.check_ident_token(cx, UnderMacro(false), ident);
+        if ident.name.as_str().starts_with('\'') {
+            self.check_ident_token(cx, UnderMacro(false), ident.without_first_quote(), "'");
+        } else {
+            self.check_ident_token(cx, UnderMacro(false), ident, "");
+        }
     }
 }