API: Float64Index.astype(datetimelike) (pandas-dev#49660)

* API: Float64Index.astype(datetimelike)

* GH ref

Author: jbrockmendel

doc/source/whatsnew/v2.0.0.rst

@@ -329,6 +329,7 @@ Other API changes
 - Default value of ``dtype`` in :func:`get_dummies` is changed to ``bool`` from ``uint8`` (:issue:`45848`)
 - :meth:`DataFrame.astype`, :meth:`Series.astype`, and :meth:`DatetimeIndex.astype` casting datetime64 data to any of "datetime64[s]", "datetime64[ms]", "datetime64[us]" will return an object with the given resolution instead of coercing back to "datetime64[ns]" (:issue:`48928`)
 - :meth:`DataFrame.astype`, :meth:`Series.astype`, and :meth:`DatetimeIndex.astype` casting timedelta64 data to any of "timedelta64[s]", "timedelta64[ms]", "timedelta64[us]" will return an object with the given resolution instead of coercing to "float64" dtype (:issue:`48963`)
+- :meth:`Index.astype` now allows casting from ``float64`` dtype to datetime-like dtypes, matching :class:`Series` behavior (:issue:`49660`)
 - Passing data with dtype of "timedelta64[s]", "timedelta64[ms]", or "timedelta64[us]" to :class:`TimedeltaIndex`, :class:`Series`, or :class:`DataFrame` constructors will now retain that dtype instead of casting to "timedelta64[ns]"; timedelta64 data with lower resolution will be cast to the lowest supported resolution "timedelta64[s]" (:issue:`49014`)
 - Passing ``dtype`` of "timedelta64[s]", "timedelta64[ms]", or "timedelta64[us]" to :class:`TimedeltaIndex`, :class:`Series`, or :class:`DataFrame` constructors will now retain that dtype instead of casting to "timedelta64[ns]"; passing a dtype with lower resolution for :class:`Series` or :class:`DataFrame` will be cast to the lowest supported resolution "timedelta64[s]" (:issue:`49014`)
 - Passing a ``np.datetime64`` object with non-nanosecond resolution to :class:`Timestamp` will retain the input resolution if it is "s", "ms", or "ns"; otherwise it will be cast to the closest supported resolution (:issue:`49008`)

pandas/core/arrays/interval.py

@@ -919,6 +919,15 @@ def astype(self, dtype, copy: bool = True):
             if dtype == self.dtype:
                 return self.copy() if copy else self
 
+            if is_float_dtype(self.dtype.subtype) and needs_i8_conversion(
+                dtype.subtype
+            ):
+                # This is allowed on the Index.astype but we disallow it here
+                msg = (
+                    f"Cannot convert {self.dtype} to {dtype}; subtypes are incompatible"
+                )
+                raise TypeError(msg)
+
             # need to cast to different subtype
             try:
                 # We need to use Index rules for astype to prevent casting

pandas/core/indexes/base.py

@@ -1016,14 +1016,6 @@ def astype(self, dtype, copy: bool = True):
             with rewrite_exception(type(values).__name__, type(self).__name__):
                 new_values = values.astype(dtype, copy=copy)
 
-        elif is_float_dtype(self.dtype) and needs_i8_conversion(dtype):
-            # NB: this must come before the ExtensionDtype check below
-            # TODO: this differs from Series behavior; can/should we align them?
-            raise TypeError(
-                f"Cannot convert dtype={self.dtype} to dtype {dtype}; integer "
-                "values are required for conversion"
-            )
-
         elif isinstance(dtype, ExtensionDtype):
             cls = dtype.construct_array_type()
             # Note: for RangeIndex and CategoricalDtype self vs self._values

pandas/tests/arithmetic/test_interval.py

@@ -133,18 +133,8 @@ def test_compare_scalar_interval_mixed_closed(self, op, closed, other_closed):
         expected = self.elementwise_comparison(op, interval_array, other)
         tm.assert_numpy_array_equal(result, expected)
 
-    def test_compare_scalar_na(
-        self, op, interval_array, nulls_fixture, box_with_array, request
-    ):
+    def test_compare_scalar_na(self, op, interval_array, nulls_fixture, box_with_array):
         box = box_with_array
-
-        if box is pd.DataFrame:
-            if interval_array.dtype.subtype.kind not in "iuf":
-                mark = pytest.mark.xfail(
-                    reason="raises on DataFrame.transpose (would be fixed by EA2D)"
-                )
-                request.node.add_marker(mark)
-
         obj = tm.box_expected(interval_array, box)
         result = op(obj, nulls_fixture)
 

pandas/tests/indexes/numeric/test_astype.py

@@ -1,11 +1,11 @@
-import re
-
 import numpy as np
 import pytest
 
-from pandas.core.dtypes.common import pandas_dtype
-
-from pandas import Index
+from pandas import (
+    Index,
+    to_datetime,
+    to_timedelta,
+)
 import pandas._testing as tm
 
 
@@ -66,15 +66,22 @@ def test_astype_float64_to_float_dtype(self, dtype):
         tm.assert_index_equal(result, expected, exact=True)
 
     @pytest.mark.parametrize("dtype", ["M8[ns]", "m8[ns]"])
-    def test_cannot_cast_to_datetimelike(self, dtype):
+    def test_astype_float_to_datetimelike(self, dtype):
+        # GH#49660 pre-2.0 Index.astype from floating to M8/m8/Period raised,
+        #  inconsistent with Series.astype
         idx = Index([0, 1.1, 2], dtype=np.float64)
 
-        msg = (
-            f"Cannot convert dtype=float64 to dtype {pandas_dtype(dtype)}; "
-            f"integer values are required for conversion"
-        )
-        with pytest.raises(TypeError, match=re.escape(msg)):
-            idx.astype(dtype)
+        result = idx.astype(dtype)
+        if dtype[0] == "M":
+            expected = to_datetime(idx.values)
+        else:
+            expected = to_timedelta(idx.values)
+        tm.assert_index_equal(result, expected)
+
+        # check that we match Series behavior
+        result = idx.to_series().set_axis(range(3)).astype(dtype)
+        expected = expected.to_series().set_axis(range(3))
+        tm.assert_series_equal(result, expected)
 
     @pytest.mark.parametrize("dtype", [int, "int16", "int32", "int64"])
     @pytest.mark.parametrize("non_finite", [np.inf, np.nan])