REF: mix PeriodPseudoDtype into PeriodDtype (#34590)

Author: jbrockmendel

pandas/_libs/tslibs/__init__.py

@@ -1,4 +1,5 @@
 __all__ = [
+    "dtypes",
     "localize_pydatetime",
     "NaT",
     "NaTType",
@@ -17,7 +18,7 @@
     "to_offset",
 ]
 
-
+from . import dtypes  # type: ignore
 from .conversion import localize_pydatetime
 from .nattype import NaT, NaTType, iNaT, is_null_datetimelike, nat_strings
 from .np_datetime import OutOfBoundsDatetime

pandas/_libs/tslibs/dtypes.pxd

@@ -50,7 +50,9 @@ cdef enum PeriodDtypeCode:
     U = 11000       # Microsecondly
     N = 12000       # Nanosecondly
 
+    UNDEFINED = -10_000
 
-cdef class PeriodPseudoDtype:
+
+cdef class PeriodDtypeBase:
     cdef readonly:
         PeriodDtypeCode dtype_code

pandas/_libs/tslibs/dtypes.pyx

@@ -2,7 +2,7 @@
 # originals
 
 
-cdef class PeriodPseudoDtype:
+cdef class PeriodDtypeBase:
     """
     Similar to an actual dtype, this contains all of the information
     describing a PeriodDtype in an integer code.
@@ -14,9 +14,9 @@ cdef class PeriodPseudoDtype:
         self.dtype_code = code
 
     def __eq__(self, other):
-        if not isinstance(other, PeriodPseudoDtype):
+        if not isinstance(other, PeriodDtypeBase):
             return False
-        if not isinstance(self, PeriodPseudoDtype):
+        if not isinstance(self, PeriodDtypeBase):
             # cython semantics, this is a reversed op
             return False
         return self.dtype_code == other.dtype_code

pandas/_libs/tslibs/offsets.pyx

@@ -2529,12 +2529,10 @@ cdef class Week(SingleConstructorOffset):
         -------
         result : DatetimeIndex
         """
-        from .frequencies import get_freq_code  # TODO: avoid circular import
-
         i8other = dtindex.asi8
         off = (i8other % DAY_NANOS).view("timedelta64[ns]")
 
-        base, mult = get_freq_code(self.freqstr)
+        base = self._period_dtype_code
         base_period = dtindex.to_period(base)
 
         if self.n > 0:

pandas/_libs/tslibs/period.pyx

@@ -56,7 +56,7 @@ from pandas._libs.tslibs.ccalendar cimport (
 )
 from pandas._libs.tslibs.ccalendar cimport c_MONTH_NUMBERS
 
-from pandas._libs.tslibs.dtypes cimport PeriodPseudoDtype
+from pandas._libs.tslibs.dtypes cimport PeriodDtypeBase
 
 from pandas._libs.tslibs.frequencies cimport (
     attrname_to_abbrevs,
@@ -1514,7 +1514,7 @@ cdef class _Period:
 
     cdef readonly:
         int64_t ordinal
-        PeriodPseudoDtype _dtype
+        PeriodDtypeBase _dtype
         BaseOffset freq
 
     def __cinit__(self, int64_t ordinal, BaseOffset freq):
@@ -1523,7 +1523,7 @@ cdef class _Period:
         # Note: this is more performant than PeriodDtype.from_date_offset(freq)
         #  because from_date_offset cannot be made a cdef method (until cython
         #  supported cdef classmethods)
-        self._dtype = PeriodPseudoDtype(freq._period_dtype_code)
+        self._dtype = PeriodDtypeBase(freq._period_dtype_code)
 
     @classmethod
     def _maybe_convert_freq(cls, object freq) -> BaseOffset:
@@ -2460,7 +2460,7 @@ class Period(_Period):
             raise ValueError(msg)
 
         if ordinal is None:
-            base, mult = get_freq_code(freq)
+            base, _ = get_freq_code(freq)
             ordinal = period_ordinal(dt.year, dt.month, dt.day,
                                      dt.hour, dt.minute, dt.second,
                                      dt.microsecond, 0, base)

pandas/core/arrays/period.py

@@ -52,7 +52,7 @@
 
 def _field_accessor(name: str, docstring=None):
     def f(self):
-        base, _ = libfrequencies.get_freq_code(self.freq)
+        base = self.freq._period_dtype_code
         result = get_period_field_arr(name, self.asi8, base)
         return result
 
@@ -440,12 +440,12 @@ def to_timestamp(self, freq=None, how="start"):
                 return (self + self.freq).to_timestamp(how="start") - adjust
 
         if freq is None:
-            base, mult = libfrequencies.get_freq_code(self.freq)
+            base = self.freq._period_dtype_code
             freq = libfrequencies.get_to_timestamp_base(base)
         else:
             freq = Period._maybe_convert_freq(freq)
 
-        base, mult = libfrequencies.get_freq_code(freq)
+        base, _ = libfrequencies.get_freq_code(freq)
         new_data = self.asfreq(freq, how=how)
 
         new_data = libperiod.periodarr_to_dt64arr(new_data.asi8, base)
@@ -523,14 +523,14 @@ def asfreq(self, freq=None, how: str = "E") -> "PeriodArray":
 
         freq = Period._maybe_convert_freq(freq)
 
-        base1, mult1 = libfrequencies.get_freq_code(self.freq)
-        base2, mult2 = libfrequencies.get_freq_code(freq)
+        base1 = self.freq._period_dtype_code
+        base2 = freq._period_dtype_code
 
         asi8 = self.asi8
-        # mult1 can't be negative or 0
+        # self.freq.n can't be negative or 0
         end = how == "E"
         if end:
-            ordinal = asi8 + mult1 - 1
+            ordinal = asi8 + self.freq.n - 1
         else:
             ordinal = asi8
 
@@ -950,7 +950,7 @@ def dt64arr_to_periodarr(data, freq, tz=None):
     if isinstance(data, (ABCIndexClass, ABCSeries)):
         data = data._values
 
-    base, mult = libfrequencies.get_freq_code(freq)
+    base = freq._period_dtype_code
     return libperiod.dt64arr_to_periodarr(data.view("i8"), base, tz), freq
 
 

pandas/core/dtypes/dtypes.py

@@ -20,7 +20,7 @@
 import pytz
 
 from pandas._libs.interval import Interval
-from pandas._libs.tslibs import NaT, Period, Timestamp, timezones, to_offset
+from pandas._libs.tslibs import NaT, Period, Timestamp, dtypes, timezones, to_offset
 from pandas._libs.tslibs.offsets import BaseOffset
 from pandas._typing import DtypeObj, Ordered
 
@@ -848,7 +848,7 @@ def __setstate__(self, state) -> None:
 
 
 @register_extension_dtype
-class PeriodDtype(PandasExtensionDtype):
+class PeriodDtype(dtypes.PeriodDtypeBase, PandasExtensionDtype):
     """
     An ExtensionDtype for Period data.
 
@@ -896,7 +896,8 @@ def __new__(cls, freq=None):
 
         elif freq is None:
             # empty constructor for pickle compat
-            u = object.__new__(cls)
+            # -10_000 corresponds to PeriodDtypeCode.UNDEFINED
+            u = dtypes.PeriodDtypeBase.__new__(cls, -10_000)
             u._freq = None
             return u
 
@@ -906,11 +907,15 @@ def __new__(cls, freq=None):
         try:
             return cls._cache[freq.freqstr]
         except KeyError:
-            u = object.__new__(cls)
+            dtype_code = freq._period_dtype_code
+            u = dtypes.PeriodDtypeBase.__new__(cls, dtype_code)
             u._freq = freq
             cls._cache[freq.freqstr] = u
             return u
 
+    def __reduce__(self):
+        return type(self), (self.freq,)
+
     @property
     def freq(self):
         """
@@ -977,7 +982,7 @@ def __eq__(self, other: Any) -> bool:
         return isinstance(other, PeriodDtype) and self.freq == other.freq
 
     def __setstate__(self, state):
-        # for pickle compat. __get_state__ is defined in the
+        # for pickle compat. __getstate__ is defined in the
         # PandasExtensionDtype superclass and uses the public properties to
         # pickle -> need to set the settable private ones here (see GH26067)
         self._freq = state["freq"]

pandas/plotting/_matplotlib/timeseries.py

@@ -6,7 +6,7 @@
 import numpy as np
 
 from pandas._libs.tslibs import Period, to_offset
-from pandas._libs.tslibs.frequencies import FreqGroup, base_and_stride, get_freq_code
+from pandas._libs.tslibs.frequencies import FreqGroup, base_and_stride
 from pandas._typing import FrameOrSeriesUnion
 
 from pandas.core.dtypes.generic import (
@@ -213,7 +213,7 @@ def _use_dynamic_x(ax, data: "FrameOrSeriesUnion") -> bool:
 
     # FIXME: hack this for 0.10.1, creating more technical debt...sigh
     if isinstance(data.index, ABCDatetimeIndex):
-        base = get_freq_code(freq)[0]
+        base = to_offset(freq)._period_dtype_code
         x = data.index
         if base <= FreqGroup.FR_DAY:
             return x[:1].is_normalized

pandas/tests/dtypes/test_dtypes.py

@@ -67,7 +67,11 @@ def test_pickle(self, dtype):
 
         # force back to the cache
         result = tm.round_trip_pickle(dtype)
-        assert not len(dtype._cache)
+        if not isinstance(dtype, PeriodDtype):
+            # Because PeriodDtype has a cython class as a base class,
+            #  it has different pickle semantics, and its cache is re-populated
+            #  on un-pickling.
+            assert not len(dtype._cache)
         assert result == dtype