feat: support HTML entities in JSX text/attributes

JSX text and attributes support HTML character references (a.k.a.
entities), and don't support ECMAScript string escape sequences.

Although the [spec] calls it "historical" and threatens to change it,
it _is_ in the spec, and the spec is pretty stable at this point.

In changing this, I landed back on an idea that @maxbrunsfeld suggested
in a [PR review] some time ago: having separate `string` and
`jsx_string` nodes, and aliasing `jsx_string` to `string` for consumers'
convenience. At that time, having two different node types was deemed
unnecessary, but this adds a second, more substantive difference between
the two, so I've brought the idea back, and stopped allowing invalid
newlines in JS string literals, which is invalid in both JS and TS.

[spec]: https://facebook.github.io/jsx/#sec-jsx-string-characters
[PR review]: #140 (comment)

grammar.js

@@ -607,10 +607,15 @@ module.exports = grammar({
 
     // Should not contain new lines and should not start or end with a space
     jsx_text: _ => choice(
-      /[^{}<>\n ]([^{}<>\n]*[^{}<>\n ])?/,
+      /[^{}<>\n& ]([^{}<>\n&]*[^{}<>\n& ])?/,
       /\/\/[^\n]*/,
     ),
 
+    // An entity can be named, numeric (decimal), or numeric (hexadecimal). The
+    // longest entity name is 29 characters long, and the HTML spec says that
+    // no more will ever be added.
+    html_character_reference: _ => /&(#([xX][0-9a-fA-F]{1,6}|[0-9]{1,5})|[A-Za-z]{1,30});/,
+
     jsx_expression: $ => seq(
       '{',
       optional(choice(
@@ -623,6 +628,7 @@ module.exports = grammar({
 
     _jsx_child: $ => choice(
       $.jsx_text,
+      $.html_character_reference,
       $._jsx_element,
       $.jsx_expression,
     ),
@@ -682,8 +688,36 @@ module.exports = grammar({
       )),
     ),
 
+    _jsx_string: $ => choice(
+      seq(
+        '"',
+        repeat(choice(
+          alias($.unescaped_double_jsx_string_fragment, $.string_fragment),
+          $.html_character_reference,
+        )),
+        '"',
+      ),
+      seq(
+        '\'',
+        repeat(choice(
+          alias($.unescaped_single_jsx_string_fragment, $.string_fragment),
+          $.html_character_reference,
+        )),
+        '\'',
+      ),
+    ),
+
+    // Workaround to https://github.com/tree-sitter/tree-sitter/issues/1156
+    // We give names to the token() constructs containing a regexp
+    // so as to obtain a node in the CST.
+    //
+    unescaped_double_jsx_string_fragment: _ => token.immediate(prec(1, /[^"&]+/)),
+
+    // same here
+    unescaped_single_jsx_string_fragment: _ => token.immediate(prec(1, /[^'&]+/)),
+
     _jsx_attribute_value: $ => choice(
-      $.string,
+      alias($._jsx_string, $.string),
       $.jsx_expression,
       $._jsx_element,
     ),
@@ -909,12 +943,6 @@ module.exports = grammar({
     // Primitives
     //
 
-    // Here we tolerate unescaped newlines in double-quoted and
-    // single-quoted string literals.
-    // This is legal in typescript as jsx/tsx attribute values (as of
-    // 2020), and perhaps will be valid in javascript as well in the
-    // future.
-    //
     string: $ => choice(
       seq(
         '"',
@@ -938,10 +966,10 @@ module.exports = grammar({
     // We give names to the token() constructs containing a regexp
     // so as to obtain a node in the CST.
     //
-    unescaped_double_string_fragment: _ => token.immediate(prec(1, /[^"\\]+/)),
+    unescaped_double_string_fragment: _ => token.immediate(prec(1, /[^"\\\r\n]+/)),
 
     // same here
-    unescaped_single_string_fragment: _ => token.immediate(prec(1, /[^'\\]+/)),
+    unescaped_single_string_fragment: _ => token.immediate(prec(1, /[^'\\\r\n]+/)),
 
     escape_sequence: _ => token.immediate(seq(
       '\\',

test/corpus/literals.txt

@@ -108,22 +108,6 @@ world';
   (expression_statement
     (string (string_fragment) (escape_sequence) (string_fragment))))
 
-============================================================
-Non-standard unescaped newlines legal in TSX attributes
-============================================================
-
-"hello
-world";
-
-'hello
-world';
-
----
-
-(program
-  (expression_statement (string (string_fragment)))
-  (expression_statement (string (string_fragment))))
-
 =========================================================
 JSX strings with unescaped newlines for TSX attributes
 =========================================================
@@ -151,3 +135,37 @@ JSX strings with unescaped newlines for TSX attributes
        (jsx_attribute (property_identifier) (string (string_fragment))))
      (jsx_closing_element
        (identifier)))))
+
+===============================================
+JSX with HTML character references (entities)
+===============================================
+
+<a>foo &nbsp; bar</a>;
+
+<abbr title="foo &nbsp; \n bar">foo</abbr>;
+
+----
+
+(program
+	(expression_statement
+		(jsx_element
+			(jsx_opening_element
+				(identifier))
+			(jsx_text)
+			(html_character_reference)
+			(jsx_text)
+			(jsx_closing_element
+				(identifier))))
+	(expression_statement
+		(jsx_element
+			(jsx_opening_element
+				(identifier)
+			(jsx_attribute
+				(property_identifier)
+				(string
+					(string_fragment)
+					(html_character_reference)
+					(string_fragment))))
+			(jsx_text)
+			(jsx_closing_element
+				(identifier)))))