BUG: pandas.DataFrame.last doesn't respect time zone #52131

doc/source/whatsnew/v2.0.1.rst

@@ -46,6 +46,7 @@ Other
 - :class:`DataFrame` created from empty dicts had :attr:`~DataFrame.columns`  of dtype ``object``. It is now a :class:`RangeIndex` (:issue:`52404`)
 - :class:`Series` created from empty dicts had :attr:`~Series.index`  of dtype ``object``. It is now a :class:`RangeIndex` (:issue:`52404`)
 - Implemented :meth:`Series.str.split` and :meth:`Series.str.rsplit` for :class:`ArrowDtype` with ``pyarrow.string`` (:issue:`52401`)
+- Implemented :meth:`_Timestamp._ensure_components` and :meth:`_Timestamp.components` timestamps.pyx and :meth:`Series.dt.components` in datetimes.py
 
 .. ---------------------------------------------------------------------------
 .. _whatsnew_201.contributors:

pandas/_libs/tslibs/timestamps.pxd

@@ -23,6 +23,8 @@ cdef class _Timestamp(ABCTimestamp):
     cdef readonly:
         int64_t _value, nanosecond, year
         NPY_DATETIMEUNIT _creso
+        bint _is_populated  # are my components populated
+        int64_t _y, _month, _d, _h, _m, _s, _us, _ns
 
     cdef bint _get_start_end_field(self, str field, freq)
     cdef _get_date_name_field(self, str field, object locale)
@@ -34,3 +36,4 @@ cdef class _Timestamp(ABCTimestamp):
                                          int op) except -1
     cdef bint _compare_mismatched_resos(_Timestamp self, _Timestamp other, int op)
     cdef _Timestamp _as_creso(_Timestamp self, NPY_DATETIMEUNIT creso, bint round_ok=*)
+    cdef _ensure_components(_Timestamp self)

pandas/_libs/tslibs/timestamps.pyx

@@ -7,6 +7,7 @@ construction requirements, we need to do object instantiation in python
 shadows the python class, where we do any heavy lifting.
 """
 
+import collections
 import warnings
 
 cimport cython
@@ -128,6 +129,21 @@ from pandas._libs.tslibs.tzconversion cimport (
 _zero_time = dt_time(0, 0)
 _no_input = object()
 
+# components named tuple
+Components = collections.namedtuple(
+        "Components",
+        [
+            "year",
+            "month",
+            "day",
+            "hour",
+            "minute",
+            "second",
+            "microsecond",
+            "nanosecond",
+        ]
+)
+
 # ----------------------------------------------------------------------
 
 
@@ -227,7 +243,6 @@ class MinMaxReso:
 # ----------------------------------------------------------------------
 
 cdef class _Timestamp(ABCTimestamp):
-
     # higher than np.ndarray and np.matrix
     __array_priority__ = 100
     dayofweek = _Timestamp.day_of_week
@@ -568,6 +583,37 @@ cdef class _Timestamp(ABCTimestamp):
             return type(self)(other) - self
         return NotImplemented
 
+    cdef _ensure_components(_Timestamp self):
+        """
+            compute the components
+        """
+
+        if self._is_populated:
+            return
+
+        cdef:
+            npy_datetimestruct dts
+
+        pandas_datetime_to_datetimestruct(self._value, self._creso, &dts)
+        self._y = dts.year
+        self._month = dts.month
+        self._d = dts.day
+        self._h = dts.hour
+        self._m = dts.min
+        self._s = dts.sec
+        self._us = dts.us
+        self._ns = dts.ps // 1000
+
+        self._is_populated = 1
+
+    @property
+    def components(self):
+        self._ensure_components()
+        return Components(
+            self._y, self._month, self._d,
+            self._h, self._m, self._s, self._us, self._ns
+            )
+
     # -----------------------------------------------------------------
 
     cdef int64_t _maybe_convert_value_to_local(self):

pandas/core/generic.py

@@ -9145,9 +9145,19 @@ def last(self, offset) -> Self:
 
         offset = to_offset(offset)
 
-        start_date = self.index[-1] - offset
-        start = self.index.searchsorted(start_date, side="right")
-        return self.iloc[start:]
+        if not isinstance(offset, Tick) and offset.is_on_offset(self.index[-1]):
+            # GH#29623 if first value is end of period, remove offset with n = 1
+            #  before adding the real offset
+            start_date = start = self.index[-1] - offset.base - offset
+        else:
+            start_date = start = self.index[-1] - offset
+
+        # Tick-like, e.g. 3 weeks
+        if isinstance(offset, Tick) and start_date in self.index:
+            start = self.index.searchsorted(start_date, side="right")
+            return self.iloc[:start]
+
+        return self.loc[:start]
 
     @final
     def rank(

pandas/tests/scalar/test_nat.py

@@ -174,7 +174,7 @@ def test_nat_iso_format(get_nat):
 @pytest.mark.parametrize(
     "klass,expected",
     [
-        (Timestamp, ["normalize", "to_julian_date", "to_period", "unit"]),
+        (Timestamp, ["components", "normalize", "to_julian_date", "to_period", "unit"]),
         (
             Timedelta,
             [

pandas/tests/scalar/timestamp/test_timestamp.py

@@ -94,6 +94,40 @@ def test_fields(self, attr, expected, tz):
         assert isinstance(result, int)
         assert result == expected
 
+        # components
+        tzstr = "dateutil/usr/share/zoneinfo/America/Chicago"
+        ts = Timestamp(
+            year=2013,
+            month=11,
+            day=3,
+            hour=1,
+            minute=0,
+            fold=1,
+            second=32,
+            microsecond=3,
+            nanosecond=7,
+            tz=tzstr,
+        ).components
+
+        assert ts.year == 2013
+        assert ts.month == 11
+        assert ts.day == 3
+        assert ts.hour == 1
+        assert ts.minute == 0
+        assert ts.second == 32
+        assert ts.microsecond == 3
+        assert ts.nanosecond == 7
+
+        tzstr = "dateutil/usr/share/zoneinfo/America/Detroit"
+        ts = Timestamp(
+            year=2023, month=4, day=14, hour=9, minute=53, fold=1, tz=tzstr
+        ).components
+        assert ts.year == 2023
+        assert ts.month == 4
+        assert ts.day == 14
+        assert ts.hour == 9
+        assert ts.minute == 53
+
     @pytest.mark.parametrize("tz", [None, "US/Eastern"])
     def test_millisecond_raises(self, tz):
         ts = Timestamp("2014-12-31 23:59:00", tz=tz)