BUG: assert_attr_equal with numpy nat or pd.NA

doc/source/whatsnew/v1.3.0.rst

@@ -432,7 +432,7 @@ Other
 - Bug in :class:`Index` constructor sometimes silently ignorning a specified ``dtype`` (:issue:`38879`)
 - Bug in constructing a :class:`Series` from a list and a :class:`PandasDtype` (:issue:`39357`)
 - Bug in :class:`Styler` which caused CSS to duplicate on multiple renders. (:issue:`39395`)
--
+- Bug in :func:`pandas.testing.assert_series_equal`, :func:`pandas.testing.assert_frame_equal`, :func:`pandas.testing.assert_index_equal` and :func:`pandas.testing.assert_extension_array_equal` incorrectly raising when an attribute has an unrecognized NA type (:issue:`39461`)
 
 .. ---------------------------------------------------------------------------
 

pandas/_testing/asserters.py

@@ -459,13 +459,24 @@ def assert_attr_equal(attr: str, left, right, obj: str = "Attributes"):
     ):
         # np.nan
         return True
+    elif (
+        isinstance(left_attr, (np.datetime64, np.timedelta64))
+        and isinstance(right_attr, (np.datetime64, np.timedelta64))
+        and type(left_attr) is type(right_attr)
+        and np.isnat(left_attr)
+        and np.isnat(right_attr)
+    ):
+        # np.datetime64("nat") or np.timedelta64("nat")
+        return True
 
     try:
         result = left_attr == right_attr
     except TypeError:
         # datetimetz on rhs may raise TypeError
         result = False
-    if not isinstance(result, bool):
+    if (left_attr is pd.NA) ^ (right_attr is pd.NA):
+        result = False
+    elif not isinstance(result, bool):
         result = result.all()
 
     if result:

pandas/core/algorithms.py

@@ -715,6 +715,8 @@ def factorize(
         values, dtype = _ensure_data(values)
 
         if original.dtype.kind in ["m", "M"]:
+            # Note: factorize_array will cast NaT bc it has a __int__
+            #  method, but will not cast the more-correct dtype.type("nat")
             na_value = iNaT
         else:
             na_value = None

pandas/core/arrays/sparse/array.py

@@ -382,6 +382,8 @@ def __init__(
                         stacklevel=2,
                     )
                     data = np.asarray(data, dtype="datetime64[ns]")
+                    if fill_value is NaT:
+                        fill_value = np.datetime64("NaT", "ns")
                 data = np.asarray(data)
             sparse_values, sparse_index, fill_value = make_sparse(
                 data, kind=kind, fill_value=fill_value, dtype=dtype

pandas/core/dtypes/missing.py

@@ -559,14 +559,14 @@ def na_value_for_dtype(dtype, compat: bool = True):
     >>> na_value_for_dtype(np.dtype('bool'))
     False
     >>> na_value_for_dtype(np.dtype('datetime64[ns]'))
-    NaT
+    numpy.datetime64('NaT')
     """
     dtype = pandas_dtype(dtype)
 
     if is_extension_array_dtype(dtype):
         return dtype.na_value
-    if needs_i8_conversion(dtype):
-        return NaT
+    elif needs_i8_conversion(dtype):
+        return dtype.type("NaT", "ns")
     elif is_float_dtype(dtype):
         return np.nan
     elif is_integer_dtype(dtype):

pandas/tests/arrays/sparse/test_dtype.py

@@ -14,8 +14,8 @@
         ("float", np.nan),
         ("bool", False),
         ("object", np.nan),
-        ("datetime64[ns]", pd.NaT),
-        ("timedelta64[ns]", pd.NaT),
+        ("datetime64[ns]", np.datetime64("NaT", "ns")),
+        ("timedelta64[ns]", np.timedelta64("NaT", "ns")),
     ],
 )
 def test_inferred_dtype(dtype, fill_value):

pandas/tests/dtypes/test_missing.py

@@ -472,8 +472,8 @@ def test_array_equivalent_nested():
     "dtype, na_value",
     [
         # Datetime-like
-        (np.dtype("M8[ns]"), NaT),
-        (np.dtype("m8[ns]"), NaT),
+        (np.dtype("M8[ns]"), np.datetime64("NaT", "ns")),
+        (np.dtype("m8[ns]"), np.timedelta64("NaT", "ns")),
         (DatetimeTZDtype.construct_from_string("datetime64[ns, US/Eastern]"), NaT),
         (PeriodDtype("M"), NaT),
         # Integer
@@ -499,7 +499,11 @@ def test_array_equivalent_nested():
 )
 def test_na_value_for_dtype(dtype, na_value):
     result = na_value_for_dtype(dtype)
-    assert result is na_value
+    # identify check doesnt work for datetime64/timedelta64("NaT") bc they
+    #  are not singletons
+    assert result is na_value or (
+        isna(result) and isna(na_value) and type(result) is type(na_value)
+    )
 
 
 class TestNAObj:

pandas/tests/util/test_assert_attr_equal.py

@@ -0,0 +1,30 @@
+from types import SimpleNamespace
+
+import pytest
+
+from pandas.core.dtypes.common import is_float
+
+import pandas._testing as tm
+
+
+def test_assert_attr_equal(nulls_fixture):
+    obj = SimpleNamespace()
+    obj.na_value = nulls_fixture
+    assert tm.assert_attr_equal("na_value", obj, obj)
+
+
+def test_assert_attr_equal_different_nulls(nulls_fixture, nulls_fixture2):
+    obj = SimpleNamespace()
+    obj.na_value = nulls_fixture
+
+    obj2 = SimpleNamespace()
+    obj2.na_value = nulls_fixture2
+
+    if nulls_fixture is nulls_fixture2:
+        assert tm.assert_attr_equal("na_value", obj, obj2)
+    elif is_float(nulls_fixture) and is_float(nulls_fixture2):
+        # we consider float("nan") and np.float64("nan") to be equivalent
+        assert tm.assert_attr_equal("na_value", obj, obj2)
+    else:
+        with pytest.raises(AssertionError, match='"na_value" are different'):
+            tm.assert_attr_equal("na_value", obj, obj2)