Merge pull request #18417 from ChayimFriedman2/hash-string

fix: Correctly handle `#""` in edition <2024

crates/parser/src/lexed_str.rs

@@ -39,7 +39,9 @@ impl<'a> LexedStr<'a> {
             conv.offset = shebang_len;
         };
 
-        for token in rustc_lexer::tokenize(&text[conv.offset..]) {
+        // Re-create the tokenizer from scratch every token because `GuardedStrPrefix` is one token in the lexer
+        // but we want to split it to two in edition <2024.
+        while let Some(token) = rustc_lexer::tokenize(&text[conv.offset..]).next() {
             let token_text = &text[conv.offset..][..token.len as usize];
 
             conv.extend_token(&token.kind, token_text);
@@ -158,7 +160,7 @@ impl<'a> Converter<'a> {
         }
     }
 
-    fn extend_token(&mut self, kind: &rustc_lexer::TokenKind, token_text: &str) {
+    fn extend_token(&mut self, kind: &rustc_lexer::TokenKind, mut token_text: &str) {
         // A note on an intended tradeoff:
         // We drop some useful information here (see patterns with double dots `..`)
         // Storing that info in `SyntaxKind` is not possible due to its layout requirements of
@@ -189,10 +191,15 @@ impl<'a> Converter<'a> {
                 rustc_lexer::TokenKind::RawIdent => IDENT,
 
                 rustc_lexer::TokenKind::GuardedStrPrefix if self.edition.at_least_2024() => {
+                    // FIXME: rustc does something better for recovery.
                     err = "Invalid string literal (reserved syntax)";
                     ERROR
                 }
-                rustc_lexer::TokenKind::GuardedStrPrefix => POUND,
+                rustc_lexer::TokenKind::GuardedStrPrefix => {
+                    // The token is `#"` or `##`, split it into two.
+                    token_text = &token_text[1..];
+                    POUND
+                }
 
                 rustc_lexer::TokenKind::Literal { kind, .. } => {
                     self.extend_literal(token_text.len(), kind);

crates/parser/src/tests.rs

@@ -15,11 +15,20 @@ use crate::{Edition, LexedStr, TopEntryPoint};
 #[path = "../test_data/generated/runner.rs"]
 mod runner;
 
+fn infer_edition(file_path: &Path) -> Edition {
+    let file_content = std::fs::read_to_string(file_path).unwrap();
+    if let Some(edition) = file_content.strip_prefix("//@ edition: ") {
+        edition[..4].parse().expect("invalid edition directive")
+    } else {
+        Edition::CURRENT
+    }
+}
+
 #[test]
 fn lex_ok() {
     for case in TestCase::list("lexer/ok") {
         let _guard = stdx::panic_context::enter(format!("{:?}", case.rs));
-        let actual = lex(&case.text);
+        let actual = lex(&case.text, infer_edition(&case.rs));
         expect_file![case.rast].assert_eq(&actual)
     }
 }
@@ -28,13 +37,13 @@ fn lex_ok() {
 fn lex_err() {
     for case in TestCase::list("lexer/err") {
         let _guard = stdx::panic_context::enter(format!("{:?}", case.rs));
-        let actual = lex(&case.text);
+        let actual = lex(&case.text, infer_edition(&case.rs));
         expect_file![case.rast].assert_eq(&actual)
     }
 }
 
-fn lex(text: &str) -> String {
-    let lexed = LexedStr::new(Edition::CURRENT, text);
+fn lex(text: &str, edition: Edition) -> String {
+    let lexed = LexedStr::new(edition, text);
 
     let mut res = String::new();
     for i in 0..lexed.len() {

crates/parser/test_data/lexer/ok/guarded_str_prefix_edition_2021.rast

@@ -0,0 +1,4 @@
+COMMENT "//@ edition: 2021"
+WHITESPACE "\n\n"
+POUND "#"
+STRING "\"foo\""

crates/parser/test_data/lexer/ok/guarded_str_prefix_edition_2021.rs

@@ -0,0 +1,3 @@
+//@ edition: 2021
+
+#"foo"