Implement .dt.total_seconds (rapidsai#17659)

Fixes: rapidsai#16802 

This PR implements `.dt.total_seconds`

Authors:
  - GALI PREM SAGAR (https://github.com/galipremsagar)

Approvers:
  - Matthew Roeschke (https://github.com/mroeschke)

URL: rapidsai#17659

Author: galipremsagar

python/cudf/cudf/core/column/timedelta.py

@@ -1,9 +1,10 @@
-# Copyright (c) 2020-2024, NVIDIA CORPORATION.
+# Copyright (c) 2020-2025, NVIDIA CORPORATION.
 
 from __future__ import annotations
 
 import datetime
 import functools
+import math
 from typing import TYPE_CHECKING, cast
 
 import numpy as np
@@ -263,7 +264,15 @@ def time_unit(self) -> str:
         return np.datetime_data(self.dtype)[0]
 
     def total_seconds(self) -> ColumnBase:
-        raise NotImplementedError("total_seconds is currently not implemented")
+        conversion = _unit_to_nanoseconds_conversion[self.time_unit] / 1e9
+        # Typecast to decimal128 to avoid floating point precision issues
+        # https://github.com/rapidsai/cudf/issues/17664
+        return (
+            (self.astype("int64") * conversion)
+            .astype(cudf.Decimal128Dtype(38, 9))
+            .round(decimals=abs(int(math.log10(conversion))))
+            .astype("float64")
+        )
 
     def ceil(self, freq: str) -> ColumnBase:
         raise NotImplementedError("ceil is currently not implemented")

python/cudf/cudf/core/index.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2018-2024, NVIDIA CORPORATION.
+# Copyright (c) 2018-2025, NVIDIA CORPORATION.
 
 from __future__ import annotations
 
@@ -842,22 +842,22 @@ def sort_values(
     @_performance_tracking
     def _gather(self, gather_map, nullify=False, check_bounds=True):
         gather_map = cudf.core.column.as_column(gather_map)
-        return cudf.Index._from_column(
+        return Index._from_column(
             self._column.take(gather_map, nullify, check_bounds),
             name=self.name,
         )
 
     @_performance_tracking
     def _apply_boolean_mask(self, boolean_mask):
-        return cudf.Index._from_column(
+        return Index._from_column(
             self._column.apply_boolean_mask(boolean_mask), name=self.name
         )
 
     def repeat(self, repeats, axis=None):
         return self._as_int_index().repeat(repeats, axis)
 
     def _split(self, splits):
-        return cudf.Index._from_column(
+        return Index._from_column(
             self._as_int_index()._split(splits), name=self.name
         )
 
@@ -1657,7 +1657,7 @@ def _clean_nulls_from_index(self) -> Index:
                 if isinstance(self, (DatetimeIndex, TimedeltaIndex))
                 else str(cudf.NA)
             )
-            return cudf.Index._from_column(
+            return Index._from_column(
                 self._column.astype("str").fillna(fill_value),
                 name=self.name,
             )
@@ -2964,13 +2964,13 @@ def median(self, *, skipna: bool = True, axis: int | None = 0):
     def std(self, *, skipna: bool = True, axis: int | None = 0, ddof: int = 1):
         return self._column.std(skipna=skipna, ddof=ddof)
 
-    def total_seconds(self) -> cupy.ndarray:
+    def total_seconds(self) -> Index:
         """
         Return total duration of each element expressed in seconds.
 
         This method is currently not implemented.
         """
-        return self._column.total_seconds().values
+        return Index._from_column(self._column.total_seconds(), name=self.name)
 
     def ceil(self, freq: str) -> Self:
         """

python/cudf/cudf/core/series.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2018-2024, NVIDIA CORPORATION.
+# Copyright (c) 2018-2025, NVIDIA CORPORATION.
 
 from __future__ import annotations
 
@@ -5183,6 +5183,66 @@ def components(self) -> cudf.DataFrame:
             ca, index=self.series.index
         )
 
+    def total_seconds(self) -> Series:
+        """
+        Return total duration of each element expressed in seconds.
+
+        This method is available directly on TimedeltaIndex
+        and on Series containing timedelta values under the ``.dt`` namespace.
+
+        Returns
+        -------
+        Index or Series
+            When the calling object is a TimedeltaIndex,
+            the return type is an Index with a float64 dtype. When the calling object
+            is a Series, the return type is Series of type `float64` whose
+            index is the same as the original.
+
+        See Also
+        --------
+        datetime.timedelta.total_seconds : Standard library version
+            of this method.
+        TimedeltaIndex.components : Return a DataFrame with components of
+            each Timedelta.
+
+        Examples
+        --------
+        **Series**
+
+        >>> import cudf
+        >>> import pandas as pd
+        >>> import numpy as np
+        >>> s = cudf.Series(pd.to_timedelta(np.arange(5), unit="D"))
+        >>> s
+        0    0 days 00:00:00
+        1    1 days 00:00:00
+        2    2 days 00:00:00
+        3    3 days 00:00:00
+        4    4 days 00:00:00
+        dtype: timedelta64[ns]
+
+        >>> s.dt.total_seconds()
+        0         0.0
+        1     86400.0
+        2    172800.0
+        3    259200.0
+        4    345600.0
+        dtype: float64
+
+        **TimedeltaIndex**
+
+        >>> idx = cudf.from_pandas(pd.to_timedelta(np.arange(5), unit="D"))
+        >>> idx
+        TimedeltaIndex(['0 days', '1 days', '2 days', '3 days', '4 days'],
+                       dtype='timedelta64[ns]', freq=None)
+
+        >>> idx.total_seconds()
+        Index([0.0, 86400.0, 172800.0, 259200.0, 345600.0], dtype='float64')
+        """
+        return self._return_result_like_self(
+            self.series._column.total_seconds()
+        )
+
 
 @_performance_tracking
 def _align_indices(series_list, how="outer", allow_non_unique=False):

python/cudf/cudf/tests/test_timedelta.py

@@ -1,4 +1,4 @@
-# Copyright (c) 2020-2024, NVIDIA CORPORATION.
+# Copyright (c) 2020-2025, NVIDIA CORPORATION.
 
 import datetime
 import operator
@@ -1506,3 +1506,25 @@ def test_tdi_unit():
     result = pd_tdi.unit
     expected = cudf_tdi.unit
     assert result == expected
+
+
+@pytest.mark.parametrize("data", _TIMEDELTA_DATA)
+@pytest.mark.parametrize("dtype", utils.TIMEDELTA_TYPES)
+def test_timedelta_series_total_seconds(data, dtype):
+    gsr = cudf.Series(data, dtype=dtype)
+    psr = gsr.to_pandas()
+
+    expected = psr.dt.total_seconds()
+    actual = gsr.dt.total_seconds()
+    assert_eq(expected, actual)
+
+
+@pytest.mark.parametrize("data", _TIMEDELTA_DATA)
+@pytest.mark.parametrize("dtype", utils.TIMEDELTA_TYPES)
+def test_timedelta_index_total_seconds(request, data, dtype):
+    gi = cudf.Index(data, dtype=dtype)
+    pi = gi.to_pandas()
+
+    expected = pi.total_seconds()
+    actual = gi.total_seconds()
+    assert_eq(expected, actual)