Rollup merge of #122217 - estebank:issue-119685, r=fmease

Handle str literals written with `'` lexed as lifetime

Given `'hello world'` and `'1 str', provide a structured suggestion for a valid string literal:

```
error[E0762]: unterminated character literal
  --> $DIR/lex-bad-str-literal-as-char-3.rs:2:26
   |
LL |     println!('hello world');
   |                          ^^^^
   |
help: if you meant to write a `str` literal, use double quotes
   |
LL |     println!("hello world");
   |              ~           ~
```
```
error[E0762]: unterminated character literal
  --> $DIR/lex-bad-str-literal-as-char-1.rs:2:20
   |
LL |     println!('1 + 1');
   |                    ^^^^
   |
help: if you meant to write a `str` literal, use double quotes
   |
LL |     println!("1 + 1");
   |              ~     ~
```

Fix #119685.

Author: matthiaskrgr

Diff for: compiler/rustc_infer/messages.ftl

@@ -169,7 +169,7 @@ infer_lifetime_param_suggestion_elided = each elided lifetime in input position
 
 infer_meant_byte_literal = if you meant to write a byte literal, prefix with `b`
 infer_meant_char_literal = if you meant to write a `char` literal, use single quotes
-infer_meant_str_literal = if you meant to write a `str` literal, use double quotes
+infer_meant_str_literal = if you meant to write a string literal, use double quotes
 infer_mismatched_static_lifetime = incompatible lifetime on type
 infer_more_targeted = {$has_param_name ->
     [true] `{$param_name}`

Diff for: compiler/rustc_infer/src/errors/mod.rs

@@ -1339,15 +1339,12 @@ pub enum TypeErrorAdditionalDiags {
         span: Span,
         code: String,
     },
-    #[suggestion(
-        infer_meant_str_literal,
-        code = "\"{code}\"",
-        applicability = "machine-applicable"
-    )]
+    #[multipart_suggestion(infer_meant_str_literal, applicability = "machine-applicable")]
     MeantStrLiteral {
-        #[primary_span]
-        span: Span,
-        code: String,
+        #[suggestion_part(code = "\"")]
+        start: Span,
+        #[suggestion_part(code = "\"")]
+        end: Span,
     },
     #[suggestion(
         infer_consider_specifying_length,

Diff for: compiler/rustc_infer/src/infer/error_reporting/mod.rs

@@ -2079,16 +2079,10 @@ impl<'tcx> TypeErrCtxt<'_, 'tcx> {
                 // If a string was expected and the found expression is a character literal,
                 // perhaps the user meant to write `"s"` to specify a string literal.
                 (ty::Ref(_, r, _), ty::Char) if r.is_str() => {
-                    if let Ok(code) = self.tcx.sess().source_map().span_to_snippet(span) {
-                        if let Some(code) =
-                            code.strip_prefix('\'').and_then(|s| s.strip_suffix('\''))
-                        {
-                            suggestions.push(TypeErrorAdditionalDiags::MeantStrLiteral {
-                                span,
-                                code: escape_literal(code),
-                            })
-                        }
-                    }
+                    suggestions.push(TypeErrorAdditionalDiags::MeantStrLiteral {
+                        start: span.with_hi(span.lo() + BytePos(1)),
+                        end: span.with_lo(span.hi() - BytePos(1)),
+                    })
                 }
                 // For code `if Some(..) = expr `, the type mismatch may be expected `bool` but found `()`,
                 // we try to suggest to add the missing `let` for `if let Some(..) = expr`

Diff for: compiler/rustc_lexer/src/cursor.rs

@@ -46,7 +46,7 @@ impl<'a> Cursor<'a> {
     /// If requested position doesn't exist, `EOF_CHAR` is returned.
     /// However, getting `EOF_CHAR` doesn't always mean actual end of file,
     /// it should be checked with `is_eof` method.
-    pub(crate) fn first(&self) -> char {
+    pub fn first(&self) -> char {
         // `.next()` optimizes better than `.nth(0)`
         self.chars.clone().next().unwrap_or(EOF_CHAR)
     }
@@ -59,6 +59,15 @@ impl<'a> Cursor<'a> {
         iter.next().unwrap_or(EOF_CHAR)
     }
 
+    /// Peeks the third symbol from the input stream without consuming it.
+    pub fn third(&self) -> char {
+        // `.next()` optimizes better than `.nth(1)`
+        let mut iter = self.chars.clone();
+        iter.next();
+        iter.next();
+        iter.next().unwrap_or(EOF_CHAR)
+    }
+
     /// Checks if there is nothing more to consume.
     pub(crate) fn is_eof(&self) -> bool {
         self.chars.as_str().is_empty()

Diff for: compiler/rustc_parse/messages.ftl

@@ -568,7 +568,7 @@ parse_more_than_one_char = character literal may only contain one codepoint
     .remove_non = consider removing the non-printing characters
     .use_double_quotes = if you meant to write a {$is_byte ->
         [true] byte string
-        *[false] `str`
+        *[false] string
         } literal, use double quotes
 
 parse_multiple_skipped_lines = multiple lines skipped by escaped newline
@@ -833,6 +833,7 @@ parse_unknown_prefix = prefix `{$prefix}` is unknown
     .label = unknown prefix
     .note =  prefixed identifiers and literals are reserved since Rust 2021
     .suggestion_br = use `br` for a raw byte string
+    .suggestion_str = if you meant to write a string literal, use double quotes
     .suggestion_whitespace = consider inserting whitespace here
 
 parse_unknown_start_of_token = unknown start of token: {$escaped}

Diff for: compiler/rustc_parse/src/errors.rs

@@ -1987,6 +1987,17 @@ pub enum UnknownPrefixSugg {
         style = "verbose"
     )]
     Whitespace(#[primary_span] Span),
+    #[multipart_suggestion(
+        parse_suggestion_str,
+        applicability = "maybe-incorrect",
+        style = "verbose"
+    )]
+    MeantStr {
+        #[suggestion_part(code = "\"")]
+        start: Span,
+        #[suggestion_part(code = "\"")]
+        end: Span,
+    },
 }
 
 #[derive(Diagnostic)]
@@ -2198,12 +2209,21 @@ pub enum MoreThanOneCharSugg {
         ch: String,
     },
     #[suggestion(parse_use_double_quotes, code = "{sugg}", applicability = "machine-applicable")]
-    Quotes {
+    QuotesFull {
         #[primary_span]
         span: Span,
         is_byte: bool,
         sugg: String,
     },
+    #[multipart_suggestion(parse_use_double_quotes, applicability = "machine-applicable")]
+    Quotes {
+        #[suggestion_part(code = "{prefix}\"")]
+        start: Span,
+        #[suggestion_part(code = "\"")]
+        end: Span,
+        is_byte: bool,
+        prefix: &'static str,
+    },
 }
 
 #[derive(Subdiagnostic)]

Diff for: compiler/rustc_parse/src/lexer/mod.rs

@@ -63,6 +63,7 @@ pub(crate) fn parse_token_trees<'psess, 'src>(
         cursor,
         override_span,
         nbsp_is_whitespace: false,
+        last_lifetime: None,
     };
     let (stream, res, unmatched_delims) =
         tokentrees::TokenTreesReader::parse_all_token_trees(string_reader);
@@ -105,6 +106,10 @@ struct StringReader<'psess, 'src> {
     /// in this file, it's safe to treat further occurrences of the non-breaking
     /// space character as whitespace.
     nbsp_is_whitespace: bool,
+
+    /// Track the `Span` for the leading `'` of the last lifetime. Used for
+    /// diagnostics to detect possible typo where `"` was meant.
+    last_lifetime: Option<Span>,
 }
 
 impl<'psess, 'src> StringReader<'psess, 'src> {
@@ -130,6 +135,18 @@ impl<'psess, 'src> StringReader<'psess, 'src> {
 
             debug!("next_token: {:?}({:?})", token.kind, self.str_from(start));
 
+            if let rustc_lexer::TokenKind::Semi
+            | rustc_lexer::TokenKind::LineComment { .. }
+            | rustc_lexer::TokenKind::BlockComment { .. }
+            | rustc_lexer::TokenKind::CloseParen
+            | rustc_lexer::TokenKind::CloseBrace
+            | rustc_lexer::TokenKind::CloseBracket = token.kind
+            {
+                // Heuristic: we assume that it is unlikely we're dealing with an unterminated
+                // string surrounded by single quotes.
+                self.last_lifetime = None;
+            }
+
             // Now "cook" the token, converting the simple `rustc_lexer::TokenKind` enum into a
             // rich `rustc_ast::TokenKind`. This turns strings into interned symbols and runs
             // additional validation.
@@ -247,6 +264,7 @@ impl<'psess, 'src> StringReader<'psess, 'src> {
                     // expansion purposes. See #12512 for the gory details of why
                     // this is necessary.
                     let lifetime_name = self.str_from(start);
+                    self.last_lifetime = Some(self.mk_sp(start, start + BytePos(1)));
                     if starts_with_number {
                         let span = self.mk_sp(start, self.pos);
                         self.dcx().struct_err("lifetimes cannot start with a number")
@@ -395,10 +413,21 @@ impl<'psess, 'src> StringReader<'psess, 'src> {
         match kind {
             rustc_lexer::LiteralKind::Char { terminated } => {
                 if !terminated {
-                    self.dcx()
+                    let mut err = self
+                        .dcx()
                         .struct_span_fatal(self.mk_sp(start, end), "unterminated character literal")
-                        .with_code(E0762)
-                        .emit()
+                        .with_code(E0762);
+                    if let Some(lt_sp) = self.last_lifetime {
+                        err.multipart_suggestion(
+                            "if you meant to write a string literal, use double quotes",
+                            vec![
+                                (lt_sp, "\"".to_string()),
+                                (self.mk_sp(start, start + BytePos(1)), "\"".to_string()),
+                            ],
+                            Applicability::MaybeIncorrect,
+                        );
+                    }
+                    err.emit()
                 }
                 self.cook_unicode(token::Char, Mode::Char, start, end, 1, 1) // ' '
             }
@@ -669,15 +698,33 @@ impl<'psess, 'src> StringReader<'psess, 'src> {
         let expn_data = prefix_span.ctxt().outer_expn_data();
 
         if expn_data.edition >= Edition::Edition2021 {
+            let mut silence = false;
             // In Rust 2021, this is a hard error.
             let sugg = if prefix == "rb" {
                 Some(errors::UnknownPrefixSugg::UseBr(prefix_span))
             } else if expn_data.is_root() {
-                Some(errors::UnknownPrefixSugg::Whitespace(prefix_span.shrink_to_hi()))
+                if self.cursor.first() == '\''
+                    && let Some(start) = self.last_lifetime
+                    && self.cursor.third() != '\''
+                {
+                    // An "unclosed `char`" error will be emitted already, silence redundant error.
+                    silence = true;
+                    Some(errors::UnknownPrefixSugg::MeantStr {
+                        start,
+                        end: self.mk_sp(self.pos, self.pos + BytePos(1)),
+                    })
+                } else {
+                    Some(errors::UnknownPrefixSugg::Whitespace(prefix_span.shrink_to_hi()))
+                }
             } else {
                 None
             };
-            self.dcx().emit_err(errors::UnknownPrefix { span: prefix_span, prefix, sugg });
+            let err = errors::UnknownPrefix { span: prefix_span, prefix, sugg };
+            if silence {
+                self.dcx().create_err(err).delay_as_bug();
+            } else {
+                self.dcx().emit_err(err);
+            }
         } else {
             // Before Rust 2021, only emit a lint for migration.
             self.psess.buffer_lint_with_diagnostic(

Diff for: compiler/rustc_parse/src/lexer/unescape_error_reporting.rs

@@ -95,11 +95,21 @@ pub(crate) fn emit_unescape_error(
                     }
                     escaped.push(c);
                 }
-                let sugg = format!("{prefix}\"{escaped}\"");
-                MoreThanOneCharSugg::Quotes {
-                    span: full_lit_span,
-                    is_byte: mode == Mode::Byte,
-                    sugg,
+                if escaped.len() != lit.len() || full_lit_span.is_empty() {
+                    let sugg = format!("{prefix}\"{escaped}\"");
+                    MoreThanOneCharSugg::QuotesFull {
+                        span: full_lit_span,
+                        is_byte: mode == Mode::Byte,
+                        sugg,
+                    }
+                } else {
+                    MoreThanOneCharSugg::Quotes {
+                        start: full_lit_span
+                            .with_hi(full_lit_span.lo() + BytePos((prefix.len() + 1) as u32)),
+                        end: full_lit_span.with_lo(full_lit_span.hi() - BytePos(1)),
+                        is_byte: mode == Mode::Byte,
+                        prefix,
+                    }
                 }
             });
             dcx.emit_err(UnescapeError::MoreThanOneChar {

Diff for: tests/ui/inference/str-as-char.stderr

@@ -4,7 +4,7 @@ error: character literal may only contain one codepoint
 LL |     let _: &str = '"""';
    |                   ^^^^^
    |
-help: if you meant to write a `str` literal, use double quotes
+help: if you meant to write a string literal, use double quotes
    |
 LL |     let _: &str = "\"\"\"";
    |                   ~~~~~~~~
@@ -15,18 +15,18 @@ error: character literal may only contain one codepoint
 LL |     let _: &str = '\"\"\"';
    |                   ^^^^^^^^
    |
-help: if you meant to write a `str` literal, use double quotes
+help: if you meant to write a string literal, use double quotes
    |
 LL |     let _: &str = "\"\"\"";
-   |                   ~~~~~~~~
+   |                   ~      ~
 
 error: character literal may only contain one codepoint
   --> $DIR/str-as-char.rs:10:19
    |
 LL |     let _: &str = '"\"\"\\"\\"';
    |                   ^^^^^^^^^^^^^^^^^
    |
-help: if you meant to write a `str` literal, use double quotes
+help: if you meant to write a string literal, use double quotes
    |
 LL |     let _: &str = "\"\"\\"\\"\\\"";
    |                   ~~~~~~~~~~~~~~~~~~~~
@@ -39,10 +39,10 @@ LL |     let _: &str = 'a';
    |            |
    |            expected due to this
    |
-help: if you meant to write a `str` literal, use double quotes
+help: if you meant to write a string literal, use double quotes
    |
 LL |     let _: &str = "a";
-   |                   ~~~
+   |                   ~ ~
 
 error: aborting due to 4 previous errors
 

Diff for: tests/ui/issues/issue-23589.stderr

@@ -15,10 +15,10 @@ error[E0308]: mismatched types
 LL |     let v: Vec(&str) = vec!['1', '2'];
    |                             ^^^ expected `&str`, found `char`
    |
-help: if you meant to write a `str` literal, use double quotes
+help: if you meant to write a string literal, use double quotes
    |
 LL |     let v: Vec(&str) = vec!["1", '2'];
-   |                             ~~~
+   |                             ~ ~
 
 error: aborting due to 2 previous errors
 

Diff for: tests/ui/lexer/lex-bad-char-literals-2.stderr

@@ -4,10 +4,10 @@ error: character literal may only contain one codepoint
 LL |     'nope'
    |     ^^^^^^
    |
-help: if you meant to write a `str` literal, use double quotes
+help: if you meant to write a string literal, use double quotes
    |
 LL |     "nope"
-   |     ~~~~~~
+   |     ~    ~
 
 error: aborting due to 1 previous error
 

Diff for: tests/ui/lexer/lex-bad-char-literals-3.stderr

@@ -4,21 +4,21 @@ error: character literal may only contain one codepoint
 LL | static c: char = '●●';
    |                  ^^^^
    |
-help: if you meant to write a `str` literal, use double quotes
+help: if you meant to write a string literal, use double quotes
    |
 LL | static c: char = "●●";
-   |                  ~~~~
+   |                  ~  ~
 
 error: character literal may only contain one codepoint
   --> $DIR/lex-bad-char-literals-3.rs:5:20
    |
 LL |     let ch: &str = '●●';
    |                    ^^^^
    |
-help: if you meant to write a `str` literal, use double quotes
+help: if you meant to write a string literal, use double quotes
    |
 LL |     let ch: &str = "●●";
-   |                    ~~~~
+   |                    ~  ~
 
 error: aborting due to 2 previous errors
 

Diff for: tests/ui/lexer/lex-bad-char-literals-5.stderr

@@ -4,21 +4,21 @@ error: character literal may only contain one codepoint
 LL | static c: char = '\x10\x10';
    |                  ^^^^^^^^^^
    |
-help: if you meant to write a `str` literal, use double quotes
+help: if you meant to write a string literal, use double quotes
    |
 LL | static c: char = "\x10\x10";
-   |                  ~~~~~~~~~~
+   |                  ~        ~
 
 error: character literal may only contain one codepoint
   --> $DIR/lex-bad-char-literals-5.rs:5:20
    |
 LL |     let ch: &str = '\x10\x10';
    |                    ^^^^^^^^^^
    |
-help: if you meant to write a `str` literal, use double quotes
+help: if you meant to write a string literal, use double quotes
    |
 LL |     let ch: &str = "\x10\x10";
-   |                    ~~~~~~~~~~
+   |                    ~        ~
 
 error: aborting due to 2 previous errors