Fixed logical op bug for numpy string arrays

Author: etong11

pandas/core/ops/array_ops.py

@@ -435,6 +435,13 @@ def fill_bool(x, left=None):
     rvalues = right
 
     if should_extension_dispatch(lvalues, rvalues):
+        # Must cast if logical op between a boolean array and numpy-backed string array
+        if ((lvalues.dtype == np.bool_ and rvalues.dtype == "string[python]")
+            or (lvalues.dtype == "string[python]" and rvalues.dtype == np.bool_)
+        ):
+            lvalues = lvalues.astype(bool)
+            rvalues = rvalues.astype(bool)
+            
         # Call the method on lvalues
         res_values = op(lvalues, rvalues)
 

pandas/tests/arrays/string_/test_string.py

@@ -766,4 +766,28 @@ def test_xor_pyarrow_string(dtype):
         ser2 = pd.Series(["", "b"], dtype=dtype)
         result = ser1 ^ ser2
         expected = pd.Series([False, True], dtype=bool)
-        tm.assert_series_equal(result, expected)
+        tm.assert_series_equal(result, expected)
+
+@pytest.mark.parametrize("dtype", ["string[python]"])
+def test_or_numpy_string(dtype):
+    ser1 = pd.Series([False, False])
+    ser2 = pd.Series(["", "b"], dtype=dtype)
+    result = ser1 | ser2
+    expected = pd.Series([False, True], dtype=bool)
+    tm.assert_series_equal(result, expected)
+
+@pytest.mark.parametrize("dtype", ["string[python]"])
+def test_and_numpy_string(dtype):
+    ser1 = pd.Series([False, False])
+    ser2 = pd.Series(["", "b"], dtype=dtype)
+    result = ser1 & ser2
+    expected = pd.Series([False, False], dtype=bool)
+    tm.assert_series_equal(result, expected)
+
+@pytest.mark.parametrize("dtype", ["string[python]"])
+def test_xor_numpy_string(dtype):
+    ser1 = pd.Series([False, False])
+    ser2 = pd.Series(["", "b"], dtype=dtype)
+    result = ser1 ^ ser2
+    expected = pd.Series([False, True], dtype=bool)
+    tm.assert_series_equal(result, expected)