utf8: throw when encoding lone surrogates

Closes #81

Author: dchest

packages/utf8/utf8.test.ts

@@ -119,4 +119,30 @@ describe("utf8", () => {
         const dec = decode(enc);
         expect(dec).toEqual(s);
     });
+
+    it("should reject invalid UTF-16 strings with unpaired surrogates", () => {
+        // High surrogate without low surrogate
+        expect(() => encode('\ud800')).toThrowError(/invalid string/);
+        expect(() => encode('\udbff')).toThrowError(/invalid string/);
+
+        // Low surrogate without high surrogate
+        expect(() => encode('\udc00')).toThrowError(/invalid string/);
+        expect(() => encode('\udfff')).toThrowError(/invalid string/);
+
+        // High surrogate at the end of string
+        expect(() => encode('hello\ud800')).toThrowError(/invalid string/);
+
+        // Low surrogate at the beginning
+        expect(() => encode('\udc00world')).toThrowError(/invalid string/);
+
+        // Two high surrogates in a row (second one is unpaired)
+        expect(() => encode('\ud800\ud800\udc00')).toThrowError(/invalid string/);
+
+        // Low surrogate followed by high surrogate (wrong order)
+        expect(() => encode('\udc00\ud800')).toThrowError(/invalid string/);
+
+        // Valid surrogate pair should work
+        expect(() => encode('\ud800\udc00')).not.toThrow(); // U+10000
+        expect(() => encode('\udbff\udfff')).not.toThrow(); // U+10FFFF
+    });
 });

packages/utf8/utf8.ts

@@ -53,7 +53,7 @@ export function encodedLength(s: string): number {
         let c = s.charCodeAt(i);
 
         if (c >= 0xd800 && c <= 0xdbff) {
-            // surrogate pair
+            // High surrogate, must be followed by low surrogate.
             if (i === s.length - 1) {
                 throw new Error(INVALID_UTF16);
             }
@@ -63,6 +63,9 @@ export function encodedLength(s: string): number {
                 throw new Error(INVALID_UTF16);
             }
             c = ((c - 0xd800) << 10) + (c2 - 0xdc00) + 0x10000;
+        } else if (c >= 0xdc00 && c <= 0xdfff) {
+            // Low surrogate without preceding high surrogate.
+            throw new Error(INVALID_UTF16);
         }
 
         if (c < 0x80) {