fix(util): 🐛 check for grapheme cluster in sfx() computation

Author: streamich

packages/json-joy/src/util/diff/__tests__/str.spec.ts

@@ -26,9 +26,11 @@ describe('sfx()', () => {
     expect(sfx('abc', '_abc')).toEqual(3);
     expect(sfx('abc', 'abcd')).toEqual(0);
     expect(sfx('👨‍🍳', '👨‍🍳')).toEqual(5);
-    // expect(sfx('👨‍🍳', '👨‍🍳chef')).toEqual(5);
-    // expect(sfx('👨‍🍳chef', '👨‍🍳')).toEqual(5);
-    // expect(sfx('👨‍🍳👨‍🍳', '👨‍🍳')).toEqual(5);
+    expect(sfx('👨‍🍳', '👨‍🍳chef')).toEqual(0);
+    expect(sfx('👨‍🍳chef', '👨‍🍳')).toEqual(0);
+    expect(sfx('👨‍🍳', 'chef👨‍🍳')).toEqual(5);
+    expect(sfx('chef👨‍🍳', '👨‍🍳')).toEqual(5);
+    expect(sfx('👨‍🍳👨‍🍳', '👨‍🍳')).toEqual(5);
   });
 });
 
@@ -435,6 +437,18 @@ describe('Unicode edge cases', () => {
     assertPatch(nfd, nfc);
     assertPatch(`hello ${nfc}`, `hello ${nfd}`);
   });
+
+   test('handles complex emoji with ZWJ sequences', () => {
+    const chefEmoji = '👨‍🍳'; // chef emoji (man + ZWJ + cooking)
+    const src = chefEmoji;
+    const dst = 'chef' + chefEmoji;
+    const patch = normalize(diff(src, dst));
+    assertPatch(src, dst, patch);
+    expect(patch).toEqual([
+      [PATCH_OP_TYPE.INS, 'chef'],
+      [PATCH_OP_TYPE.EQL, chefEmoji],
+    ]);
+  });
 });
 
 describe('Algorithm edge cases', () => {

packages/json-joy/src/util/diff/str.ts

@@ -427,9 +427,35 @@ export const sfx = (txt1: string, txt2: string): number => {
     } else max = mid;
     mid = Math.floor((max - min) / 2 + min);
   }
-  const code = txt1.charCodeAt(txt1.length - mid);
-  const isSurrogatePairEnd = code >= 0xd800 && code <= 0xdbff;
-  if (isSurrogatePairEnd) mid--;
+  // Check if we're splitting a surrogate pair or combining character sequence
+  // We need to check the character BEFORE the matched suffix to see if we're
+  // splitting a grapheme cluster.
+  if (mid > 0 && mid < txt1.length) {
+    const boundaryPos = txt1.length - mid - 1;
+    const code = txt1.charCodeAt(boundaryPos);
+    const isHighSurrogate = code >= 0xd800 && code <= 0xdbff;
+    const isCombining = 
+      code === 0x200d || // ZWJ
+      (code >= 0xfe00 && code <= 0xfe0f) || // Variation selectors
+      (code >= 0x0300 && code <= 0x036f); // Combining diacritical marks
+    
+    if (isHighSurrogate || isCombining) {
+      // We're splitting a grapheme cluster. Walk backwards to include the full cluster.
+      mid--;
+      while (mid > 0) {
+        const pos = txt1.length - mid - 1;
+        if (pos < 0) break;
+        const prevCode = txt1.charCodeAt(pos);
+        const isPrevHighSurrogate = prevCode >= 0xd800 && prevCode <= 0xdbff;
+        const isPrevCombining = 
+          prevCode === 0x200d ||
+          (prevCode >= 0xfe00 && prevCode <= 0xfe0f) ||
+          (prevCode >= 0x0300 && prevCode <= 0x036f);
+        if (!isPrevHighSurrogate && !isPrevCombining) break;
+        mid--;
+      }
+    }
+  }
   return mid;
 };