fix: pd.to_numeric handling of datetime

Author: suzyahyah

doc/source/whatsnew/v3.0.0.rst

@@ -1191,6 +1191,7 @@ Other
 - Bug in :func:`eval` where method calls on binary operations like ``(x + y).dropna()`` would raise ``AttributeError: 'BinOp' object has no attribute 'value'`` (:issue:`61175`)
 - Bug in :func:`eval` where the names of the :class:`Series` were not preserved when using ``engine="numexpr"``. (:issue:`10239`)
 - Bug in :func:`eval` with ``engine="numexpr"`` returning unexpected result for float division. (:issue:`59736`)
+- Bug in :func:`to_numeric` for ``datetime``, :class:`Series` and ``NaT`` conversions. (:issue:`43280`)
 - Bug in :func:`to_numeric` raising ``TypeError`` when ``arg`` is a :class:`Timedelta` or :class:`Timestamp` scalar. (:issue:`59944`)
 - Bug in :func:`unique` on :class:`Index` not always returning :class:`Index` (:issue:`57043`)
 - Bug in :meth:`DataFrame.apply` raising ``RecursionError`` when passing ``func=list[int]``. (:issue:`61565`)

pandas/_libs/lib.pyi

@@ -124,6 +124,7 @@ def maybe_convert_numeric(
     na_values: set,
     convert_empty: bool = ...,
     coerce_numeric: bool = ...,
+    convert_datetime: bool = ...,
     convert_to_masked_nullable: Literal[False] = ...,
 ) -> tuple[np.ndarray, None]: ...
 @overload

pandas/_libs/lib.pyx

@@ -2300,6 +2300,7 @@ def maybe_convert_numeric(
     set na_values,
     bint convert_empty=True,
     bint coerce_numeric=False,
+    bint convert_datetime=True,
     bint convert_to_masked_nullable=False,
 ) -> tuple[np.ndarray, np.ndarray | None]:
     """
@@ -2449,6 +2450,18 @@ def maybe_convert_numeric(
         elif is_decimal(val):
             floats[i] = complexes[i] = val
             seen.float_ = True
+        elif convert_datetime and (PyDate_Check(val) or cnp.is_datetime64_object(val)):
+            # convert_datetime flag avoids conversion for base_readers
+            # PyDate_Check also includes PyDatetime_Check
+            seen.datetime_ = True
+            if val in na_values or checknull(val):
+                seen.saw_null()
+                mask[i] = 1
+                floats[i] = NaN
+            else:
+                ints[i] = np.datetime64(val).astype(int)
+                # because of pd.NaT, we may need to return in floats #GH 42380
+                floats[i] = float(ints[i])
         else:
             try:
                 floatify(val, &fval, &maybe_int)
@@ -2517,7 +2530,7 @@ def maybe_convert_numeric(
         if seen.null_ and convert_to_masked_nullable:
             return (floats, mask.view(np.bool_))
         return (floats, None)
-    elif seen.int_:
+    elif seen.int_ or seen.datetime_:
         if seen.null_ and convert_to_masked_nullable:
             if seen.uint_:
                 return (uints, mask.view(np.bool_))

pandas/core/tools/numeric.py

@@ -204,8 +204,12 @@ def to_numeric(
             return float(arg)
         if is_number(arg):
             return arg
-        if isinstance(arg, (Timedelta, Timestamp)):
+        if isinstance(arg, Timedelta):
             return arg._value
+        if isinstance(arg, Timestamp):
+            if arg.tzinfo:
+                arg = arg.tz_convert("UTC").replace(tzinfo=None)
+
         is_scalars = True
         values = np.array([arg], dtype="O")
     elif getattr(arg, "ndim", 1) > 1:
@@ -227,8 +231,6 @@ def to_numeric(
     new_mask: np.ndarray | None = None
     if is_numeric_dtype(values_dtype):
         pass
-    elif lib.is_np_dtype(values_dtype, "mM"):
-        values = values.view(np.int64)
     else:
         values = ensure_object(values)
         coerce_numeric = errors != "raise"

pandas/io/parsers/base_parser.py

@@ -490,6 +490,7 @@ def _infer_types(
                     values,
                     na_values,
                     False,
+                    convert_datetime=False,
                     convert_to_masked_nullable=non_default_dtype_backend,  # type: ignore[arg-type]
                 )
             except (ValueError, TypeError):

pandas/tests/tools/test_to_numeric.py

@@ -1,4 +1,6 @@
+from datetime import datetime
 import decimal
+from functools import partial
 
 import numpy as np
 from numpy import iinfo
@@ -902,6 +904,54 @@ def test_to_numeric_dtype_backend_error(dtype_backend):
     tm.assert_series_equal(result, expected)
 
 
+@pytest.mark.parametrize(
+    "input_value, expected, pd_type",
+    [
+        (datetime(2021, 8, 22), 1629590400000000, "scalar"),
+        (datetime(2025, 2, 21), 1740096000000000, "scalar"),
+        (pd.NaT, np.nan, "scalar"),
+        ([datetime(2021, 8, 22)], [1629590400000000], "series"),
+        ([datetime(2025, 2, 21)], [1740096000000000], "series"),
+        ([pd.NaT], [np.nan], "series"),
+        ([datetime(2021, 8, 22), pd.NaT], [float(1629590400000000), np.nan], "series"),
+        ([pd.NaT, datetime(2021, 8, 22)], [np.nan, float(1629590400000000)], "series"),
+        (
+            ["apple", 1, datetime(2021, 8, 22)],
+            [np.nan, float(1.0), float(1629590400000000)],
+            "series_coerce",
+        ),
+        ([pd.NaT], [np.nan], "series_partial"),
+        ([datetime(2025, 2, 21)], [1740096000000000], "series_partial"),
+        (
+            [pd.NaT, datetime(2025, 2, 21)],
+            [np.nan, float(1740096000000000)],
+            "series_partial",
+        ),
+    ],
+)
+def test_to_numeric_datetime(input_value, expected, pd_type):
+    """Test converting a scalar datetime to numeric."""
+    if pd_type == "scalar":
+        val = to_numeric(input_value)
+        # special handling because  Nan!=Nan
+        if pd.isna(expected):
+            assert pd.isna(val)
+        else:
+            assert val == expected
+
+    elif pd_type == "series":
+        val = to_numeric(Series(input_value))
+        tm.assert_series_equal(val, Series(expected))
+
+    elif pd_type == "series_coerce":
+        val = to_numeric(Series(input_value), errors="coerce")
+        tm.assert_series_equal(val, Series(expected))
+
+    elif pd_type == "series_partial":
+        val = Series(input_value).apply(partial(to_numeric))
+        tm.assert_series_equal(val, Series(expected))
+
+
 def test_invalid_dtype_backend():
     ser = Series([1, 2, 3])
     msg = (