API: Timestamp and Timedelta .value changing in 2.0

pandas/_libs/index.pyx

@@ -517,14 +517,16 @@ cdef class DatetimeEngine(Int64Engine):
         # NB: caller is responsible for ensuring tzawareness compat
         #  before we get here
         if scalar is NaT:
-            return NaT.value
+            return NaT._value
         elif isinstance(scalar, _Timestamp):
             if scalar._creso == self._creso:
-                return scalar.value
+                return scalar._value
             else:
                 # Note: caller is responsible for catching potential ValueError
                 #  from _as_creso
-                return (<_Timestamp>scalar)._as_creso(self._creso, round_ok=False).value
+                return (
+                    (<_Timestamp>scalar)._as_creso(self._creso, round_ok=False)._value
+                )
         raise TypeError(scalar)
 
     def __contains__(self, val: object) -> bool:
@@ -585,22 +587,24 @@ cdef class TimedeltaEngine(DatetimeEngine):
 
     cdef int64_t _unbox_scalar(self, scalar) except? -1:
         if scalar is NaT:
-            return NaT.value
+            return NaT._value
         elif isinstance(scalar, _Timedelta):
             if scalar._creso == self._creso:
-                return scalar.value
+                return scalar._value
             else:
                 # Note: caller is responsible for catching potential ValueError
                 #  from _as_creso
-                return (<_Timedelta>scalar)._as_creso(self._creso, round_ok=False).value
+                return (
+                    (<_Timedelta>scalar)._as_creso(self._creso, round_ok=False)._value
+                )
         raise TypeError(scalar)
 
 
 cdef class PeriodEngine(Int64Engine):
 
     cdef int64_t _unbox_scalar(self, scalar) except? -1:
         if scalar is NaT:
-            return scalar.value
+            return scalar._value
         if is_period_object(scalar):
             # NB: we assume that we have the correct freq here.
             return scalar.ordinal

pandas/_libs/src/ujson/python/objToJSON.c

@@ -1306,9 +1306,9 @@ char **NpyArr_encodeLabels(PyArrayObject *labels, PyObjectEncoder *enc,
             castfunc(dataptr, &nanosecVal, 1, NULL, NULL);
         } else if (PyDate_Check(item) || PyDelta_Check(item)) {
             is_datetimelike = 1;
-            if (PyObject_HasAttrString(item, "value")) {
+            if (PyObject_HasAttrString(item, "_value")) {
                 // see test_date_index_and_values for case with non-nano
-                nanosecVal = get_long_attr(item, "value");
+                nanosecVal = get_long_attr(item, "_value");
             } else {
                 if (PyDelta_Check(item)) {
                     nanosecVal = total_seconds(item) *
@@ -1554,8 +1554,8 @@ void Object_beginTypeContext(JSOBJ _obj, JSONTypeContext *tc) {
         }
         return;
     } else if (PyDelta_Check(obj)) {
-        if (PyObject_HasAttrString(obj, "value")) {
-            value = get_long_attr(obj, "value");
+        if (PyObject_HasAttrString(obj, "_value")) {
+            value = get_long_attr(obj, "_value");
         } else {
             value = total_seconds(obj) * 1000000000LL;  // nanoseconds per sec
         }

pandas/_libs/tslib.pyx

@@ -765,7 +765,7 @@ def array_to_datetime_with_tz(ndarray values, tzinfo tz):
                     # datetime64, tznaive pydatetime, int, float
                     ts = ts.tz_localize(tz)
                 ts = ts.as_unit("ns")
-                ival = ts.value
+                ival = ts._value
 
         # Analogous to: result[i] = ival
         (<int64_t*>cnp.PyArray_MultiIter_DATA(mi, 0))[0] = ival

pandas/_libs/tslibs/conversion.pyx

@@ -765,7 +765,7 @@ cdef int64_t parse_pydatetime(
             result = _ts.value
     else:
         if isinstance(val, _Timestamp):
-            result = val.as_unit("ns").value
+            result = val.as_unit("ns")._value
         else:
             result = pydatetime_to_dt64(val, dts)
             check_dts_bounds(dts)

pandas/_libs/tslibs/nattype.pxd

@@ -8,7 +8,7 @@ cdef set c_nat_strings
 
 cdef class _NaT(datetime):
     cdef readonly:
-        int64_t value
+        int64_t _value
 
 cdef _NaT c_NaT
 

pandas/_libs/tslibs/nattype.pyi

@@ -19,6 +19,7 @@ class _NatComparison:
 
 class NaTType:
     value: np.int64
+    _value: np.int64
     @property
     def asm8(self) -> np.datetime64: ...
     def to_datetime64(self) -> np.datetime64: ...

pandas/_libs/tslibs/nattype.pyx

@@ -357,7 +357,7 @@ class NaTType(_NaT):
         cdef _NaT base
 
         base = _NaT.__new__(cls, 1, 1, 1)
-        base.value = NPY_NAT
+        base._value= NPY_NAT
 
         return base
 

pandas/_libs/tslibs/offsets.pyx

@@ -183,7 +183,7 @@ def apply_wraps(func):
                     res = result.tz_localize(None)
                 else:
                     res = result
-                value = res.as_unit("ns").value
+                value = res.as_unit("ns")._value
                 result = Timestamp(value + nano)
 
         if tz is not None and result.tzinfo is None:
@@ -3469,7 +3469,7 @@ cdef class FY5253Quarter(FY5253Mixin):
         else:
             tdelta = Timedelta(0)
 
-        # Note: we always have tdelta.value >= 0
+        # Note: we always have tdelta._value>= 0
         return start, num_qtrs, tdelta
 
     @apply_wraps
@@ -3481,7 +3481,7 @@ cdef class FY5253Quarter(FY5253Mixin):
         prev_year_end, num_qtrs, tdelta = self._rollback_to_year(other)
         res = prev_year_end
         n += num_qtrs
-        if self.n <= 0 and tdelta.value > 0:
+        if self.n <= 0 and tdelta._value > 0:
             n += 1
 
         # Possible speedup by handling years first.

pandas/_libs/tslibs/strptime.pyx

@@ -113,14 +113,14 @@ cdef bint parse_today_now(str val, int64_t* iresult, bint utc):
     #  microsecond resolution
     if val == "now":
         if utc:
-            iresult[0] = Timestamp.utcnow().value * 1000
+            iresult[0] = Timestamp.utcnow()._value * 1000
         else:
             # GH#18705 make sure to_datetime("now") matches Timestamp("now")
             # Note using Timestamp.now() is faster than Timestamp("now")
-            iresult[0] = Timestamp.now().value * 1000
+            iresult[0] = Timestamp.now()._value * 1000
         return True
     elif val == "today":
-        iresult[0] = Timestamp.today().value * 1000
+        iresult[0] = Timestamp.today()._value * 1000
         return True
     return False
 
@@ -284,7 +284,7 @@ def array_strptime(
                     utc,
                 )
                 if isinstance(val, _Timestamp):
-                    iresult[i] = val.tz_localize(None).as_unit("ns").value
+                    iresult[i] = val.tz_localize(None).as_unit("ns")._value
                 else:
                     iresult[i] = pydatetime_to_dt64(val.replace(tzinfo=None), &dts)
                     check_dts_bounds(&dts)

pandas/_libs/tslibs/timedeltas.pxd

@@ -14,7 +14,7 @@ cdef bint is_any_td_scalar(object obj)
 
 cdef class _Timedelta(timedelta):
     cdef readonly:
-        int64_t value      # nanoseconds
+        int64_t _value      # nanoseconds
         bint _is_populated  # are my components populated
         int64_t _d, _h, _m, _s, _ms, _us, _ns
         NPY_DATETIMEUNIT _creso

pandas/_libs/tslibs/timedeltas.pyi

@@ -89,6 +89,7 @@ class Timedelta(timedelta):
     max: ClassVar[Timedelta]
     resolution: ClassVar[Timedelta]
     value: int  # np.int64
+    _value: int  # np.int64
     # error: "__new__" must return a class instance (got "Union[Timestamp, NaTType]")
     def __new__(  # type: ignore[misc]
         cls: type[_S],