Backport PR #54913 on branch 2.1.x (REGR: drop_duplicates raising for…

… arrow strings) (#54955)

Backport PR #54913: REGR: drop_duplicates raising for arrow strings

Co-authored-by: Patrick Hoefler <61934744+phofl@users.noreply.github.com>

doc/source/whatsnew/v2.1.1.rst

@@ -16,6 +16,7 @@ Fixed regressions
 - Fixed regression in :func:`merge` when merging over a PyArrow string index (:issue:`54894`)
 - Fixed regression in :func:`read_csv` when ``usecols`` is given and ``dtypes`` is a dict for ``engine="python"`` (:issue:`54868`)
 - Fixed regression in :meth:`DataFrame.__setitem__` raising ``AssertionError`` when setting a :class:`Series` with a partial :class:`MultiIndex` (:issue:`54875`)
+- Fixed regression in :meth:`Series.drop_duplicates` for PyArrow strings (:issue:`54904`)
 - Fixed regression in :meth:`Series.value_counts` raising for numeric data if ``bins`` was specified (:issue:`54857`)
 - Fixed regression when comparing a :class:`Series` with ``datetime64`` dtype with ``None`` (:issue:`54870`)
 

pandas/core/algorithms.py

@@ -1000,7 +1000,7 @@ def duplicated(
     duplicated : ndarray[bool]
     """
     if hasattr(values, "dtype"):
-        if isinstance(values.dtype, ArrowDtype):
+        if isinstance(values.dtype, ArrowDtype) and values.dtype.kind in "ifub":
             values = values._to_masked()  # type: ignore[union-attr]
 
         if isinstance(values.dtype, BaseMaskedDtype):

pandas/tests/series/methods/test_drop_duplicates.py

@@ -1,6 +1,7 @@
 import numpy as np
 import pytest
 
+import pandas as pd
 from pandas import (
     Categorical,
     Series,
@@ -256,3 +257,11 @@ def test_duplicated_arrow_dtype(self):
         result = ser.drop_duplicates()
         expected = Series([True, False, None], dtype="bool[pyarrow]")
         tm.assert_series_equal(result, expected)
+
+    def test_drop_duplicates_arrow_strings(self):
+        # GH#54904
+        pa = pytest.importorskip("pyarrow")
+        ser = Series(["a", "a"], dtype=pd.ArrowDtype(pa.string()))
+        result = ser.drop_duplicates()
+        expecetd = Series(["a"], dtype=pd.ArrowDtype(pa.string()))
+        tm.assert_series_equal(result, expecetd)