fix[split]: keep surrogate pairs and combining marks together for an empty separator

src/compat/_internal/regexUnicode.ts

@@ -0,0 +1,33 @@
+/**
+ * Matches a single visible symbol (grapheme cluster) in a Unicode `string`,
+ * combining surrogate pairs, combining marks, variation selectors, Fitzpatrick
+ * modifiers, regional-indicator pairs, and zero-width-joiner sequences into a
+ * single match.
+ *
+ * Ported from lodash to preserve identical behavior.
+ * @link https://github.com/lodash/lodash/blob/4.17.21-es/_unicodeSize.js
+ */
+
+const rsAstralRange = '\\ud800-\\udfff';
+const rsComboMarksRange = '\\u0300-\\u036f';
+const reComboHalfMarksRange = '\\ufe20-\\ufe2f';
+const rsComboSymbolsRange = '\\u20d0-\\u20ff';
+const rsComboRange = rsComboMarksRange + reComboHalfMarksRange + rsComboSymbolsRange;
+const rsVarRange = '\\ufe0e\\ufe0f';
+
+const rsAstral = `[${rsAstralRange}]`;
+const rsCombo = `[${rsComboRange}]`;
+const rsFitz = '\\ud83c[\\udffb-\\udfff]';
+const rsModifier = `(?:${rsCombo}|${rsFitz})`;
+const rsNonAstral = `[^${rsAstralRange}]`;
+const rsRegional = '(?:\\ud83c[\\udde6-\\uddff]){2}';
+const rsSurrPair = '[\\ud800-\\udbff][\\udc00-\\udfff]';
+const rsZWJ = '\\u200d';
+
+const reOptMod = `${rsModifier}?`;
+const rsOptVar = `[${rsVarRange}]?`;
+const rsOptJoin = `(?:${rsZWJ}(?:${[rsNonAstral, rsRegional, rsSurrPair].join('|')})${rsOptVar}${reOptMod})*`;
+const rsSeq = rsOptVar + reOptMod + rsOptJoin;
+const rsSymbol = `(?:${[`${rsNonAstral}${rsCombo}?`, rsCombo, rsRegional, rsSurrPair, rsAstral].join('|')})`;
+
+export const regexUnicode = RegExp(`${rsFitz}(?=${rsFitz})|${rsSymbol}${rsSeq}`, 'g');

src/compat/_internal/unicodeSize.ts

@@ -1,42 +1,16 @@
+import { regexUnicode } from './regexUnicode.ts';
+
 /**
- * Counts the number of visible symbols (grapheme clusters) in a Unicode `string`,
- * combining surrogate pairs, combining marks, variation selectors, Fitzpatrick
- * modifiers, regional-indicator pairs, and zero-width-joiner sequences into a
- * single symbol.
+ * Counts the number of visible symbols (grapheme clusters) in a Unicode `string`.
  *
  * Ported from lodash to preserve identical behavior.
  * @link https://github.com/lodash/lodash/blob/4.17.21-es/_unicodeSize.js
  */
-
-const rsAstralRange = '\\ud800-\\udfff';
-const rsComboMarksRange = '\\u0300-\\u036f';
-const reComboHalfMarksRange = '\\ufe20-\\ufe2f';
-const rsComboSymbolsRange = '\\u20d0-\\u20ff';
-const rsComboRange = rsComboMarksRange + reComboHalfMarksRange + rsComboSymbolsRange;
-const rsVarRange = '\\ufe0e\\ufe0f';
-
-const rsAstral = `[${rsAstralRange}]`;
-const rsCombo = `[${rsComboRange}]`;
-const rsFitz = '\\ud83c[\\udffb-\\udfff]';
-const rsModifier = `(?:${rsCombo}|${rsFitz})`;
-const rsNonAstral = `[^${rsAstralRange}]`;
-const rsRegional = '(?:\\ud83c[\\udde6-\\uddff]){2}';
-const rsSurrPair = '[\\ud800-\\udbff][\\udc00-\\udfff]';
-const rsZWJ = '\\u200d';
-
-const reOptMod = `${rsModifier}?`;
-const rsOptVar = `[${rsVarRange}]?`;
-const rsOptJoin = `(?:${rsZWJ}(?:${[rsNonAstral, rsRegional, rsSurrPair].join('|')})${rsOptVar}${reOptMod})*`;
-const rsSeq = rsOptVar + reOptMod + rsOptJoin;
-const rsSymbol = `(?:${[`${rsNonAstral}${rsCombo}?`, rsCombo, rsRegional, rsSurrPair, rsAstral].join('|')})`;
-
-const reUnicode = RegExp(`${rsFitz}(?=${rsFitz})|${rsSymbol}${rsSeq}`, 'g');
-
 export function unicodeSize(str: string): number {
   let result = 0;
-  reUnicode.lastIndex = 0;
+  regexUnicode.lastIndex = 0;
 
-  while (reUnicode.test(str)) {
+  while (regexUnicode.test(str)) {
     ++result;
   }
 

src/compat/_internal/unicodeToArray.ts

@@ -0,0 +1,11 @@
+import { regexUnicode } from './regexUnicode.ts';
+
+/**
+ * Splits a Unicode `string` into an array of its visible symbols (grapheme clusters).
+ *
+ * Ported from lodash to preserve identical behavior.
+ * @link https://github.com/lodash/lodash/blob/4.17.21-es/_unicodeToArray.js
+ */
+export function unicodeToArray(str: string): string[] {
+  return str.match(regexUnicode) || [];
+}

src/compat/string/split.spec.ts

@@ -32,4 +32,11 @@ describe('split', () => {
     const result = split('abc', 'b').join(',');
     expect(result).toBe('a,c');
   });
+
+  it('should split unicode strings into symbols when `separator` is an empty string', () => {
+    expect(split('Ame\u0301lie', '')).toEqual(['A', 'm', 'e\u0301', 'l', 'i', 'e']);
+    expect(split('a😀b', '')).toEqual(['a', '😀', 'b']);
+    expect(split('👨‍👩‍👧🇰🇷', '')).toEqual(['👨‍👩‍👧', '🇰🇷']);
+    expect(split('a😀b', '', 2)).toEqual(['a', '😀']);
+  });
 });

src/compat/string/split.ts

@@ -1,3 +1,5 @@
+import { regexMultiByte } from '../_internal/regexMultiByte.ts';
+import { unicodeToArray } from '../_internal/unicodeToArray.ts';
 import { toString } from '../util/toString.ts';
 
 /**
@@ -36,6 +38,12 @@
  */
 export function split(string: string | null | undefined, index: string | number, guard: object): string[];
 
 export function split(string: any, separator?: any, limit?: any): string[] {
-  return toString(string).split(separator as string, limit);
+  const str = toString(string);
+
+  if (separator === '' && regexMultiByte.test(str)) {
+    return unicodeToArray(str).slice(0, limit === undefined ? undefined : limit >>> 0);
+  }
+
+  return str.split(separator as string, limit);
 }