Initial support for raw lifetimes

compiler-errors · compiler-errors · commit e6bf16c47467 · 2024-08-27T20:32:59.000-04:00
diff --git a/compiler/rustc_ast/src/mut_visit.rs b/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, _ /* raw */) | token::Lifetime(name, _ /* 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) => {
diff --git a/compiler/rustc_ast/src/token.rs b/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,
         }
@@ -646,7 +646,9 @@ 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),
         }
     }
@@ -664,11 +666,11 @@ 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,
         }
     }
@@ -850,7 +852,7 @@ 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,
             },
 
diff --git a/compiler/rustc_ast/src/tokenstream.rs b/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),
diff --git a/compiler/rustc_ast_pretty/src/pprust/state.rs b/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) => {
diff --git a/compiler/rustc_expand/src/mbe/macro_parser.rs b/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
     }
diff --git a/compiler/rustc_expand/src/mbe/transcribe.rs b/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)) => {
diff --git a/compiler/rustc_expand/src/proc_macro_server.rs b/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),
diff --git a/compiler/rustc_lexer/src/lib.rs b/compiler/rustc_lexer/src/lib.rs
@@ -97,6 +97,10 @@ pub enum TokenKind {
     /// and not the separator.
     UnknownPrefixLifetime,
 
+    /// `'r`, which is parsed into `'r#lt` in edition 2021 and above. Before
+    /// edition 2021, we warn that it's a reserved syntax.
+    RawLifetimePrefix,
+
     /// 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,
@@ -683,9 +687,14 @@ impl Cursor<'_> {
             return Literal { kind, suffix_start };
         }
 
+        if self.first() == 'r' && self.second() == '#' && is_id_start(self.third()) {
+            // Eat "r" character.
+            self.bump();
+            return RawLifetimePrefix;
+        }
+
         // 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.
diff --git a/compiler/rustc_lint/messages.ftl b/compiler/rustc_lint/messages.ftl
@@ -700,6 +700,10 @@ 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
diff --git a/compiler/rustc_lint/src/context/diagnostics.rs b/compiler/rustc_lint/src/context/diagnostics.rs
@@ -171,6 +171,10 @@ pub(super) fn decorate_lint(sess: &Session, diagnostic: BuiltinLintDiag, diag: &
             }
             .decorate_lint(diag);
         }
+        BuiltinLintDiag::RawPrefix(label_span) => {
+            lints::RawPrefix { label: label_span, suggestion: label_span.shrink_to_hi() }
+                .decorate_lint(diag);
+        }
         BuiltinLintDiag::UnusedBuiltinAttribute { attr_name, macro_name, invoc_span } => {
             lints::UnusedBuiltinAttribute { invoc_span, attr_name, macro_name }.decorate_lint(diag);
         }
diff --git a/compiler/rustc_lint/src/lints.rs b/compiler/rustc_lint/src/lints.rs
@@ -2750,6 +2750,15 @@ pub(crate) struct ReservedPrefix {
     pub prefix: String,
 }
 
+#[derive(LintDiagnostic)]
+#[diag(lint_raw_prefix)]
+pub(crate) struct RawPrefix {
+    #[label]
+    pub label: Span,
+    #[suggestion(code = " ", applicability = "machine-applicable")]
+    pub suggestion: Span,
+}
+
 #[derive(LintDiagnostic)]
 #[diag(lint_unused_builtin_attribute)]
 pub(crate) struct UnusedBuiltinAttribute {
diff --git a/compiler/rustc_lint_defs/src/lib.rs b/compiler/rustc_lint_defs/src/lib.rs
@@ -623,6 +623,8 @@ pub enum BuiltinLintDiag {
     LegacyDeriveHelpers(Span),
     OrPatternsBackCompat(Span, String),
     ReservedPrefix(Span, String),
+    /// `'r#` in edition < 2021.
+    RawPrefix(Span),
     TrailingMacro(bool, Ident),
     BreakWithLabelAndLoop(Span),
     UnicodeTextFlow(Span, String),
diff --git a/compiler/rustc_parse/src/lexer/mod.rs b/compiler/rustc_parse/src/lexer/mod.rs
@@ -1,3 +1,4 @@
+use std::assert_matches::assert_matches;
 use std::ops::Range;
 
 use rustc_ast::ast::{self, AttrStyle};
@@ -13,7 +14,6 @@ use rustc_session::lint::builtin::{
 };
 use rustc_session::lint::BuiltinLintDiag;
 use rustc_session::parse::ParseSess;
-use rustc_span::edition::Edition;
 use rustc_span::symbol::Symbol;
 use rustc_span::{BytePos, Pos, Span};
 use tracing::debug;
@@ -284,7 +284,49 @@ impl<'psess, 'src> StringReader<'psess, 'src> {
                             .stash(span, StashKey::LifetimeIsChar);
                     }
                     let ident = Symbol::intern(lifetime_name);
-                    token::Lifetime(ident)
+                    token::Lifetime(ident, IdentIsRaw::No)
+                }
+                rustc_lexer::TokenKind::RawLifetimePrefix => {
+                    let prefix_span = self.mk_sp(start, start + BytePos(2));
+                    if prefix_span.at_least_rust_2021() {
+                        // #
+                        let hash_token = self.cursor.advance_token();
+                        self.pos = self.pos + BytePos(hash_token.len);
+                        assert_matches!(hash_token.kind, rustc_lexer::TokenKind::Pound);
+
+                        // ident
+                        let ident_start = self.pos;
+                        let ident_token = self.cursor.advance_token();
+                        self.pos = self.pos + BytePos(ident_token.len);
+                        match ident_token.kind {
+                            rustc_lexer::TokenKind::Ident => {}
+                            rustc_lexer::TokenKind::UnknownPrefix | rustc_lexer::TokenKind::RawIdent => {
+                                self.report_unknown_prefix(start);
+                            }
+                            _ => unreachable!(),
+                        }
+
+                        let lifetime_name_without_tick = self.str_from(ident_start);
+                        // Put the `'` back onto the lifetime name.
+                        let mut lifetime_name = String::with_capacity(lifetime_name_without_tick.len() + 1);
+                        lifetime_name.push('\'');
+                        lifetime_name += lifetime_name_without_tick;
+                        let sym = Symbol::intern(&lifetime_name);
+                        token::Lifetime(sym, IdentIsRaw::Yes)
+                    } else {
+                        // Otherwise, this is just `'r`. Warn about it though.
+                        self.psess.buffer_lint(
+                            RUST_2021_PREFIXES_INCOMPATIBLE_SYNTAX,
+                            prefix_span,
+                            ast::CRATE_NODE_ID,
+                            BuiltinLintDiag::RawPrefix(prefix_span),
+                        );
+
+                        let lifetime_name = self.str_from(start);
+                        self.last_lifetime = Some(prefix_span);
+                        let ident = Symbol::intern(lifetime_name);
+                        token::Lifetime(ident, IdentIsRaw::No)
+                    }
                 }
                 rustc_lexer::TokenKind::Semi => token::Semi,
                 rustc_lexer::TokenKind::Comma => token::Comma,
@@ -712,7 +754,7 @@ impl<'psess, 'src> StringReader<'psess, 'src> {
 
         let expn_data = prefix_span.ctxt().outer_expn_data();
 
-        if expn_data.edition >= Edition::Edition2021 {
+        if expn_data.edition.at_least_rust_2021() {
             // In Rust 2021, this is a hard error.
             let sugg = if prefix == "rb" {
                 Some(errors::UnknownPrefixSugg::UseBr(prefix_span))
diff --git a/compiler/rustc_parse/src/parser/expr.rs b/compiler/rustc_parse/src/parser/expr.rs
@@ -2050,7 +2050,7 @@ impl<'a> Parser<'a> {
         };
         // On an error path, eagerly consider a lifetime to be an unclosed character lit, if that
         // makes sense.
-        if let Some(ident) = self.token.lifetime()
+        if let Some((ident, IdentIsRaw::No)) = self.token.lifetime()
             && could_be_unclosed_char_literal(ident)
         {
             let lt = self.expect_lifetime();
@@ -2925,9 +2925,9 @@ impl<'a> Parser<'a> {
     }
 
     pub(crate) fn eat_label(&mut self) -> Option<Label> {
-        if let Some(ident) = self.token.lifetime() {
+        if let Some((ident, is_raw)) = self.token.lifetime() {
             // Disallow `'fn`, but with a better error message than `expect_lifetime`.
-            if ident.without_first_quote().is_reserved() {
+            if matches!(is_raw, IdentIsRaw::No) && ident.without_first_quote().is_reserved() {
                 self.dcx().emit_err(errors::InvalidLabel { span: ident.span, name: ident.name });
             }
 
diff --git a/compiler/rustc_parse/src/parser/mod.rs b/compiler/rustc_parse/src/parser/mod.rs
@@ -1666,7 +1666,7 @@ enum FlatToken {
 pub enum ParseNtResult {
     Tt(TokenTree),
     Ident(Ident, IdentIsRaw),
-    Lifetime(Ident),
+    Lifetime(Ident, IdentIsRaw),
 
     /// This case will eventually be removed, along with `Token::Interpolate`.
     Nt(Lrc<Nonterminal>),
diff --git a/compiler/rustc_parse/src/parser/nonterminal.rs b/compiler/rustc_parse/src/parser/nonterminal.rs
@@ -106,7 +106,7 @@ impl<'a> Parser<'a> {
                 _ => false,
             },
             NonterminalKind::Lifetime => match &token.kind {
-                token::Lifetime(_) | token::NtLifetime(..) => true,
+                token::Lifetime(..) | token::NtLifetime(..) => true,
                 _ => false,
             },
             NonterminalKind::TT | NonterminalKind::Item | NonterminalKind::Stmt => {
@@ -189,9 +189,9 @@ impl<'a> Parser<'a> {
             NonterminalKind::Lifetime => {
                 // We want to keep `'keyword` parsing, just like `keyword` is still
                 // an ident for nonterminal purposes.
-                return if let Some(ident) = self.token.lifetime() {
+                return if let Some((ident, is_raw)) = self.token.lifetime() {
                     self.bump();
-                    Ok(ParseNtResult::Lifetime(ident))
+                    Ok(ParseNtResult::Lifetime(ident, is_raw))
                 } else {
                     Err(self.dcx().create_err(UnexpectedNonterminal::Lifetime {
                         span: self.token.span,
diff --git a/compiler/rustc_parse/src/parser/pat.rs b/compiler/rustc_parse/src/parser/pat.rs
@@ -1,6 +1,6 @@
 use rustc_ast::mut_visit::{walk_pat, MutVisitor};
 use rustc_ast::ptr::P;
-use rustc_ast::token::{self, BinOpToken, Delimiter, Token};
+use rustc_ast::token::{self, BinOpToken, Delimiter, IdentIsRaw, Token};
 use rustc_ast::{
     self as ast, AttrVec, BindingMode, ByRef, Expr, ExprKind, MacCall, Mutability, Pat, PatField,
     PatFieldsRest, PatKind, Path, QSelf, RangeEnd, RangeSyntax,
@@ -544,7 +544,7 @@ impl<'a> Parser<'a> {
                     None => PatKind::Path(qself, path),
                 }
             }
-        } else if let Some(lt) = self.token.lifetime()
+        } else if let Some((lt, IdentIsRaw::No)) = self.token.lifetime()
             // In pattern position, we're totally fine with using "next token isn't colon"
             // as a heuristic. We could probably just always try to recover if it's a lifetime,
             // because we never have `'a: label {}` in a pattern position anyways, but it does
@@ -685,7 +685,7 @@ impl<'a> Parser<'a> {
     /// Parse `&pat` / `&mut pat`.
     fn parse_pat_deref(&mut self, expected: Option<Expected>) -> PResult<'a, PatKind> {
         self.expect_and()?;
-        if let Some(lifetime) = self.token.lifetime() {
+        if let Some((lifetime, _)) = self.token.lifetime() {
             self.bump(); // `'a`
 
             self.dcx().emit_err(UnexpectedLifetimeInPattern {
diff --git a/compiler/rustc_parse/src/parser/ty.rs b/compiler/rustc_parse/src/parser/ty.rs

Original file line number	Diff line number	Diff line change
`@@ -283,9 +283,9 @@ pub(super) fn transcribe<'a>(`
`283`	`283`	`let kind = token::NtIdent(ident, is_raw);`
`284`	`284`	`TokenTree::token_alone(kind, sp)`
`285`	`285`	`}`
`286`		`- MatchedSingle(ParseNtResult::Lifetime(ident)) => {`
	`286`	`+ MatchedSingle(ParseNtResult::Lifetime(ident, is_raw)) => {`
`287`	`287`	`marker.visit_span(&mut sp);`
`288`		`- let kind = token::NtLifetime(*ident);`
	`288`	`+ let kind = token::NtLifetime(ident, is_raw);`
`289`	`289`	`TokenTree::token_alone(kind, sp)`
`290`	`290`	`}`
`291`	`291`	`MatchedSingle(ParseNtResult::Nt(nt)) => {`
Original file line number	Diff line number	Diff line change
`@@ -171,6 +171,10 @@ pub(super) fn decorate_lint(sess: &Session, diagnostic: BuiltinLintDiag, diag: &`
`171`	`171`	`}`
`172`	`172`	`.decorate_lint(diag);`
`173`	`173`	`}`
	`174`	`+ BuiltinLintDiag::RawPrefix(label_span) => {`
	`175`	`+ lints::RawPrefix { label: label_span, suggestion: label_span.shrink_to_hi() }`
	`176`	`+ .decorate_lint(diag);`
	`177`	`+ }`
`174`	`178`	`BuiltinLintDiag::UnusedBuiltinAttribute { attr_name, macro_name, invoc_span } => {`
`175`	`179`	`lints::UnusedBuiltinAttribute { invoc_span, attr_name, macro_name }.decorate_lint(diag);`
`176`	`180`	`}`