BUG: Don't multiply sets during construction (#32594)

Author: dsaxton

doc/source/whatsnew/v1.1.0.rst

@@ -274,7 +274,7 @@ Conversion
 ^^^^^^^^^^
 - Bug in :class:`Series` construction from NumPy array with big-endian ``datetime64`` dtype (:issue:`29684`)
 - Bug in :class:`Timedelta` construction with large nanoseconds keyword value (:issue:`32402`)
--
+- Bug in :class:`DataFrame` construction where sets would be duplicated rather than raising (:issue:`32582`)
 
 Strings
 ^^^^^^^

pandas/core/construction.py

@@ -5,6 +5,7 @@
 These should not depend on core.internals.
 """
 
+from collections import abc
 from typing import TYPE_CHECKING, Any, Optional, Sequence, Union, cast
 
 import numpy as np
@@ -446,6 +447,8 @@ def sanitize_array(
         # GH#16804
         arr = np.arange(data.start, data.stop, data.step, dtype="int64")
         subarr = _try_cast(arr, dtype, copy, raise_cast_failure)
+    elif isinstance(data, abc.Set):
+        raise TypeError("Set type is unordered")
     else:
         subarr = _try_cast(data, dtype, copy, raise_cast_failure)
 

pandas/tests/frame/test_constructors.py

@@ -2604,3 +2604,9 @@ def test_from_2d_ndarray_with_dtype(self):
 
         expected = DataFrame(array_dim2).astype("datetime64[ns, UTC]")
         tm.assert_frame_equal(df, expected)
+
+    def test_construction_from_set_raises(self):
+        # https://github.com/pandas-dev/pandas/issues/32582
+        msg = "Set type is unordered"
+        with pytest.raises(TypeError, match=msg):
+            pd.DataFrame({"a": {1, 2, 3}})