From c76af4f338fb99a31d43bf43c8b53af37f11f567 Mon Sep 17 00:00:00 2001
From: Patrick Meredith <pmeredit@protonmail.com>
Date: Wed, 23 Sep 2026 14:42:43 -0400
Subject: [PATCH] fix: accept bare ampersands in JSX text

`<p>Shipping & Handling</p>` is valid JSX (React renders the ampersand),
but scan_jsx_text stopped at every `&` and the grammar only accepts a full
character reference there, so a bare `&` was a parse error.

The scanner now looks past an `&`: if what follows is the rest of an HTML
character reference as html_character_reference defines it (`&amp;`,
`&#38;`, `&#x26;`), the text ends before it and the reference is its own
token as before; otherwise the ampersand is text.

Fixes #366

Co-Authored-By: Claude Fable 5.1 <noreply@anthropic.com>
---
 src/scanner.c            | 47 ++++++++++++++++++++++++++++++++++++++--
 test/corpus/literals.txt | 38 ++++++++++++++++++++++++++++++++
 2 files changed, 83 insertions(+), 2 deletions(-)

diff --git a/src/scanner.c b/src/scanner.c
index 795916dd..7a87c3e5 100644
--- a/src/scanner.c
+++ b/src/scanner.c
@@ -289,6 +289,35 @@ static bool scan_html_comment(TSLexer *lexer) {
     return true;
 }
 
+// After an `&`: is what follows the rest of an HTML character reference
+// (`&amp;`, `&#38;`, `&#x26;`), as the grammar's html_character_reference
+// token defines it? Consumes the name or number either way.
+static bool scan_character_reference_tail(TSLexer *lexer) {
+    unsigned length = 0;
+    if (lexer->lookahead == '#') {
+        advance(lexer);
+        if (lexer->lookahead == 'x' || lexer->lookahead == 'X') {
+            advance(lexer);
+            while (length < 6 && iswxdigit(lexer->lookahead)) {
+                advance(lexer);
+                length++;
+            }
+        } else {
+            while (length < 5 && iswdigit(lexer->lookahead)) {
+                advance(lexer);
+                length++;
+            }
+        }
+    } else {
+        while (length < 30 && ((lexer->lookahead >= 'a' && lexer->lookahead <= 'z') ||
+                               (lexer->lookahead >= 'A' && lexer->lookahead <= 'Z'))) {
+            advance(lexer);
+            length++;
+        }
+    }
+    return length > 0 && lexer->lookahead == ';';
+}
+
 static bool scan_jsx_text(TSLexer *lexer) {
     // saw_text will be true if we see any non-whitespace content, or any whitespace content that is not a newline and
     // does not immediately follow a newline.
@@ -297,8 +326,22 @@ static bool scan_jsx_text(TSLexer *lexer) {
     // immediately follows a newline.
     bool at_newline = false;
 
+    lexer->result_symbol = JSX_TEXT;
+
     while (lexer->lookahead != 0 && lexer->lookahead != '<' && lexer->lookahead != '>' && lexer->lookahead != '{' &&
-           lexer->lookahead != '}' && lexer->lookahead != '&') {
+           lexer->lookahead != '}') {
+        if (lexer->lookahead == '&') {
+            // A character reference is its own token: the text ends before it.
+            // A bare ampersand (`Shipping & Handling`) is text.
+            lexer->mark_end(lexer);
+            advance(lexer);
+            if (scan_character_reference_tail(lexer)) {
+                return saw_text;
+            }
+            saw_text = true;
+            at_newline = false;
+            continue;
+        }
         bool is_wspace = iswspace(lexer->lookahead);
         if (lexer->lookahead == '\n') {
             at_newline = true;
@@ -326,7 +369,7 @@ static bool scan_jsx_text(TSLexer *lexer) {
         advance(lexer);
     }
 
-    lexer->result_symbol = JSX_TEXT;
+    lexer->mark_end(lexer);
     return saw_text;
 }
 
diff --git a/test/corpus/literals.txt b/test/corpus/literals.txt
index 4a617652..9df86da7 100644
--- a/test/corpus/literals.txt
+++ b/test/corpus/literals.txt
@@ -171,3 +171,41 @@ JSX with HTML character references (entities)
 			(jsx_text)
 			(jsx_closing_element
 				(identifier)))))
+=============================================
+JSX with bare ampersands in text
+=============================================
+
+<p>Shipping & Handling</p>;
+
+<p>a &b; c &#; d &#x; e &amp;</p>;
+
+<p>&</p>;
+
+----
+
+(program
+	(expression_statement
+		(jsx_element
+			(jsx_opening_element
+				(identifier))
+			(jsx_text)
+			(jsx_closing_element
+				(identifier))))
+	(expression_statement
+		(jsx_element
+			(jsx_opening_element
+				(identifier))
+			(jsx_text)
+			(html_character_reference)
+			(jsx_text)
+			(html_character_reference)
+			(jsx_closing_element
+				(identifier))))
+	(expression_statement
+		(jsx_element
+			(jsx_opening_element
+				(identifier))
+			(jsx_text)
+			(jsx_closing_element
+				(identifier)))))
+