BUG: fix Series.value_counts with sort=False returns result sorted on values for Series with string dtype (#57116)

Author: yuanx749

doc/source/whatsnew/v3.0.0.rst

@@ -152,7 +152,7 @@ Conversion
 
 Strings
 ^^^^^^^
--
+- Bug in :meth:`Series.value_counts` would not respect ``sort=False`` for series having ``string`` dtype (:issue:`55224`)
 -
 
 Interval

pandas/core/arrays/string_.py

@@ -542,7 +542,7 @@ def max(self, axis=None, skipna: bool = True, **kwargs) -> Scalar:
     def value_counts(self, dropna: bool = True) -> Series:
         from pandas.core.algorithms import value_counts_internal as value_counts
 
-        result = value_counts(self._ndarray, dropna=dropna).astype("Int64")
+        result = value_counts(self._ndarray, sort=False, dropna=dropna).astype("Int64")
         result.index = result.index.astype(self.dtype)
         return result
 

pandas/tests/arrays/string_/test_string.py

@@ -584,6 +584,19 @@ def test_value_counts_with_normalize(dtype):
     tm.assert_series_equal(result, expected)
 
 
+def test_value_counts_sort_false(dtype):
+    if getattr(dtype, "storage", "") == "pyarrow":
+        exp_dtype = "int64[pyarrow]"
+    elif getattr(dtype, "storage", "") == "pyarrow_numpy":
+        exp_dtype = "int64"
+    else:
+        exp_dtype = "Int64"
+    ser = pd.Series(["a", "b", "c", "b"], dtype=dtype)
+    result = ser.value_counts(sort=False)
+    expected = pd.Series([1, 2, 1], index=ser[:3], dtype=exp_dtype, name="count")
+    tm.assert_series_equal(result, expected)
+
+
 @pytest.mark.parametrize(
     "values, expected",
     [