Remove address violation with safe_sort (#55984)

Author: WillAyd

pandas/core/algorithms.py

@@ -1630,11 +1630,12 @@ def safe_sort(
     if use_na_sentinel:
         # take_nd is faster, but only works for na_sentinels of -1
         order2 = sorter.argsort()
-        new_codes = take_nd(order2, codes, fill_value=-1)
         if verify:
             mask = (codes < -len(values)) | (codes >= len(values))
+            codes[mask] = 0
         else:
             mask = None
+        new_codes = take_nd(order2, codes, fill_value=-1)
     else:
         reverse_indexer = np.empty(len(sorter), dtype=int)
         reverse_indexer.put(sorter, np.arange(len(sorter)))

pandas/tests/test_sorting.py

@@ -5,11 +5,6 @@
 import numpy as np
 import pytest
 
-from pandas.compat import (
-    is_ci_environment,
-    is_platform_windows,
-)
-
 from pandas import (
     NA,
     DataFrame,
@@ -409,11 +404,6 @@ def test_codes(self, verify, codes, exp_codes):
         tm.assert_numpy_array_equal(result, expected)
         tm.assert_numpy_array_equal(result_codes, expected_codes)
 
-    @pytest.mark.skipif(
-        is_platform_windows() and is_ci_environment(),
-        reason="In CI environment can crash thread with: "
-        "Windows fatal exception: access violation",
-    )
     def test_codes_out_of_bound(self):
         values = np.array([3, 1, 2, 0, 4])
         expected = np.array([0, 1, 2, 3, 4])