REF/PERF: PeriodDtype decouple from DateOffset

pandas/_libs/tslibs/dtypes.pxd

@@ -0,0 +1,56 @@
+
+cdef enum PeriodDtypeCode:
+    # Annual freqs with various fiscal year ends.
+    # eg, 2005 for A_FEB runs Mar 1, 2004 to Feb 28, 2005
+    A = 1000      # Default alias
+    A_DEC = 1000  # Annual - December year end
+    A_JAN = 1001  # Annual - January year end
+    A_FEB = 1002  # Annual - February year end
+    A_MAR = 1003  # Annual - March year end
+    A_APR = 1004  # Annual - April year end
+    A_MAY = 1005  # Annual - May year end
+    A_JUN = 1006  # Annual - June year end
+    A_JUL = 1007  # Annual - July year end
+    A_AUG = 1008  # Annual - August year end
+    A_SEP = 1009  # Annual - September year end
+    A_OCT = 1010  # Annual - October year end
+    A_NOV = 1011  # Annual - November year end
+
+    # Quarterly frequencies with various fiscal year ends.
+    # eg, Q42005 for Q_OCT runs Aug 1, 2005 to Oct 31, 2005
+    Q_DEC = 2000    # Quarterly - December year end
+    Q_JAN = 2001    # Quarterly - January year end
+    Q_FEB = 2002    # Quarterly - February year end
+    Q_MAR = 2003    # Quarterly - March year end
+    Q_APR = 2004    # Quarterly - April year end
+    Q_MAY = 2005    # Quarterly - May year end
+    Q_JUN = 2006    # Quarterly - June year end
+    Q_JUL = 2007    # Quarterly - July year end
+    Q_AUG = 2008    # Quarterly - August year end
+    Q_SEP = 2009    # Quarterly - September year end
+    Q_OCT = 2010    # Quarterly - October year end
+    Q_NOV = 2011    # Quarterly - November year end
+
+    M = 3000        # Monthly
+
+    W_SUN = 4000    # Weekly - Sunday end of week
+    W_MON = 4001    # Weekly - Monday end of week
+    W_TUE = 4002    # Weekly - Tuesday end of week
+    W_WED = 4003    # Weekly - Wednesday end of week
+    W_THU = 4004    # Weekly - Thursday end of week
+    W_FRI = 4005    # Weekly - Friday end of week
+    W_SAT = 4006    # Weekly - Saturday end of week
+
+    B = 5000        # Business days
+    D = 6000        # Daily
+    H = 7000        # Hourly
+    T = 8000        # Minutely
+    S = 9000        # Secondly
+    L = 10000       # Millisecondly
+    U = 11000       # Microsecondly
+    N = 12000       # Nanosecondly
+
+
+cdef class PeriodPseudoDtype:
+    cdef readonly:
+        PeriodDtypeCode dtype_code

pandas/_libs/tslibs/dtypes.pyx

@@ -0,0 +1,108 @@
+# period frequency constants corresponding to scikits timeseries
+# originals
+
+
+cdef class PeriodPseudoDtype:
+    """
+    Similar to an actual dtype, this contains all of the information
+    describing a PeriodDtype in an integer code.
+    """
+    # cdef readonly:
+    #    PeriodDtypeCode dtype_code
+
+    def __cinit__(self, PeriodDtypeCode code):
+        self.dtype_code = code
+
+    def __eq__(self, other):
+        if not isinstance(other, PeriodPseudoDtype):
+            return False
+        if not isinstance(self, PeriodPseudoDtype):
+            # cython semantics, this is a reversed op
+            return False
+        return self.dtype_code == other.dtype_code
+
+    @property
+    def date_offset(self):
+        """
+        Corresponding DateOffset object.
+
+        This mapping is mainly for backward-compatibility.
+        """
+        from .offsets import to_offset
+
+        freqstr = _reverse_period_code_map.get(self.dtype_code)
+        # equiv: freqstr = libfrequencies.get_freq_str(self.dtype_code)
+
+        return to_offset(freqstr)
+
+    @classmethod
+    def from_date_offset(cls, offset):
+        code = offset._period_dtype_code
+        return cls(code)
+
+
+_period_code_map = {
+    # Annual freqs with various fiscal year ends.
+    # eg, 2005 for A-FEB runs Mar 1, 2004 to Feb 28, 2005
+    "A-DEC": 1000,  # Annual - December year end
+    "A-JAN": 1001,  # Annual - January year end
+    "A-FEB": 1002,  # Annual - February year end
+    "A-MAR": 1003,  # Annual - March year end
+    "A-APR": 1004,  # Annual - April year end
+    "A-MAY": 1005,  # Annual - May year end
+    "A-JUN": 1006,  # Annual - June year end
+    "A-JUL": 1007,  # Annual - July year end
+    "A-AUG": 1008,  # Annual - August year end
+    "A-SEP": 1009,  # Annual - September year end
+    "A-OCT": 1010,  # Annual - October year end
+    "A-NOV": 1011,  # Annual - November year end
+
+    # Quarterly frequencies with various fiscal year ends.
+    # eg, Q42005 for Q-OCT runs Aug 1, 2005 to Oct 31, 2005
+    "Q-DEC": 2000,    # Quarterly - December year end
+    "Q-JAN": 2001,    # Quarterly - January year end
+    "Q-FEB": 2002,    # Quarterly - February year end
+    "Q-MAR": 2003,    # Quarterly - March year end
+    "Q-APR": 2004,    # Quarterly - April year end
+    "Q-MAY": 2005,    # Quarterly - May year end
+    "Q-JUN": 2006,    # Quarterly - June year end
+    "Q-JUL": 2007,    # Quarterly - July year end
+    "Q-AUG": 2008,    # Quarterly - August year end
+    "Q-SEP": 2009,    # Quarterly - September year end
+    "Q-OCT": 2010,    # Quarterly - October year end
+    "Q-NOV": 2011,    # Quarterly - November year end
+
+    "M": 3000,        # Monthly
+
+    "W-SUN": 4000,    # Weekly - Sunday end of week
+    "W-MON": 4001,    # Weekly - Monday end of week
+    "W-TUE": 4002,    # Weekly - Tuesday end of week
+    "W-WED": 4003,    # Weekly - Wednesday end of week
+    "W-THU": 4004,    # Weekly - Thursday end of week
+    "W-FRI": 4005,    # Weekly - Friday end of week
+    "W-SAT": 4006,    # Weekly - Saturday end of week
+
+    "B": 5000,        # Business days
+    "D": 6000,        # Daily
+    "H": 7000,        # Hourly
+    "T": 8000,        # Minutely
+    "S": 9000,        # Secondly
+    "L": 10000,       # Millisecondly
+    "U": 11000,       # Microsecondly
+    "N": 12000,       # Nanosecondly
+}
+
+_reverse_period_code_map = {
+    _period_code_map[key]: key for key in _period_code_map}
+
+# Yearly aliases; careful not to put these in _reverse_period_code_map
+_period_code_map.update({"Y" + key[1:]: _period_code_map[key]
+                         for key in _period_code_map
+                         if key.startswith("A-")})
+
+_period_code_map.update({
+    "Q": 2000,   # Quarterly - December year end (default quarterly)
+    "A": 1000,   # Annual
+    "W": 4000,   # Weekly
+    "C": 5000,   # Custom Business Day
+})

pandas/_libs/tslibs/frequencies.pyx

@@ -12,6 +12,8 @@ from pandas._libs.tslibs.offsets import (
     opattern,
 )
 
+from .dtypes import _period_code_map, _reverse_period_code_map
+
 # ---------------------------------------------------------------------
 # Period codes
 
@@ -31,73 +33,6 @@ class FreqGroup:
     FR_NS = 12000
 
 
-# period frequency constants corresponding to scikits timeseries
-# originals
-_period_code_map = {
-    # Annual freqs with various fiscal year ends.
-    # eg, 2005 for A-FEB runs Mar 1, 2004 to Feb 28, 2005
-    "A-DEC": 1000,  # Annual - December year end
-    "A-JAN": 1001,  # Annual - January year end
-    "A-FEB": 1002,  # Annual - February year end
-    "A-MAR": 1003,  # Annual - March year end
-    "A-APR": 1004,  # Annual - April year end
-    "A-MAY": 1005,  # Annual - May year end
-    "A-JUN": 1006,  # Annual - June year end
-    "A-JUL": 1007,  # Annual - July year end
-    "A-AUG": 1008,  # Annual - August year end
-    "A-SEP": 1009,  # Annual - September year end
-    "A-OCT": 1010,  # Annual - October year end
-    "A-NOV": 1011,  # Annual - November year end
-
-    # Quarterly frequencies with various fiscal year ends.
-    # eg, Q42005 for Q-OCT runs Aug 1, 2005 to Oct 31, 2005
-    "Q-DEC": 2000,    # Quarterly - December year end
-    "Q-JAN": 2001,    # Quarterly - January year end
-    "Q-FEB": 2002,    # Quarterly - February year end
-    "Q-MAR": 2003,    # Quarterly - March year end
-    "Q-APR": 2004,    # Quarterly - April year end
-    "Q-MAY": 2005,    # Quarterly - May year end
-    "Q-JUN": 2006,    # Quarterly - June year end
-    "Q-JUL": 2007,    # Quarterly - July year end
-    "Q-AUG": 2008,    # Quarterly - August year end
-    "Q-SEP": 2009,    # Quarterly - September year end
-    "Q-OCT": 2010,    # Quarterly - October year end
-    "Q-NOV": 2011,    # Quarterly - November year end
-
-    "M": 3000,        # Monthly
-
-    "W-SUN": 4000,    # Weekly - Sunday end of week
-    "W-MON": 4001,    # Weekly - Monday end of week
-    "W-TUE": 4002,    # Weekly - Tuesday end of week
-    "W-WED": 4003,    # Weekly - Wednesday end of week
-    "W-THU": 4004,    # Weekly - Thursday end of week
-    "W-FRI": 4005,    # Weekly - Friday end of week
-    "W-SAT": 4006,    # Weekly - Saturday end of week
-
-    "B": 5000,        # Business days
-    "D": 6000,        # Daily
-    "H": 7000,        # Hourly
-    "T": 8000,        # Minutely
-    "S": 9000,        # Secondly
-    "L": 10000,       # Millisecondly
-    "U": 11000,       # Microsecondly
-    "N": 12000}       # Nanosecondly
-
-
-_reverse_period_code_map = {
-    _period_code_map[key]: key for key in _period_code_map}
-
-# Yearly aliases; careful not to put these in _reverse_period_code_map
-_period_code_map.update({'Y' + key[1:]: _period_code_map[key]
-                         for key in _period_code_map
-                         if key.startswith('A-')})
-
-_period_code_map.update({
-    "Q": 2000,   # Quarterly - December year end (default quarterly)
-    "A": 1000,   # Annual
-    "W": 4000,   # Weekly
-    "C": 5000})  # Custom Business Day
-
 # Map attribute-name resolutions to resolution abbreviations
 _attrname_to_abbrevs = {
     "year": "A",

pandas/_libs/tslibs/offsets.pyx

@@ -48,6 +48,7 @@ from pandas._libs.tslibs.np_datetime cimport (
 from pandas._libs.tslibs.timezones cimport utc_pytz as UTC
 from pandas._libs.tslibs.tzconversion cimport tz_convert_single
 
+from .dtypes cimport PeriodDtypeCode
 from .timedeltas cimport delta_to_nanoseconds
 
 
@@ -892,36 +893,43 @@ cdef class Tick(SingleConstructorOffset):
 cdef class Day(Tick):
     _nanos_inc = 24 * 3600 * 1_000_000_000
     _prefix = "D"
+    _period_dtype_code = PeriodDtypeCode.D
 
 
 cdef class Hour(Tick):
     _nanos_inc = 3600 * 1_000_000_000
     _prefix = "H"
+    _period_dtype_code = PeriodDtypeCode.H
 
 
 cdef class Minute(Tick):
     _nanos_inc = 60 * 1_000_000_000
     _prefix = "T"
+    _period_dtype_code = PeriodDtypeCode.T
 
 
 cdef class Second(Tick):
     _nanos_inc = 1_000_000_000
     _prefix = "S"
+    _period_dtype_code = PeriodDtypeCode.S
 
 
 cdef class Milli(Tick):
     _nanos_inc = 1_000_000
     _prefix = "L"
+    _period_dtype_code = PeriodDtypeCode.L
 
 
 cdef class Micro(Tick):
     _nanos_inc = 1000
     _prefix = "U"
+    _period_dtype_code = PeriodDtypeCode.U
 
 
 cdef class Nano(Tick):
     _nanos_inc = 1
     _prefix = "N"
+    _period_dtype_code = PeriodDtypeCode.N
 
 
 def delta_to_tick(delta: timedelta) -> Tick:
@@ -1281,7 +1289,7 @@ cdef class BusinessDay(BusinessMixin):
     """
     DateOffset subclass representing possibly n business days.
     """
-
+    _period_dtype_code = PeriodDtypeCode.B
     _prefix = "B"
     _attributes = tuple(["n", "normalize", "offset"])
 
@@ -1945,6 +1953,15 @@ cdef class YearEnd(YearOffset):
     _prefix = "A"
     _day_opt = "end"
 
+    cdef readonly:
+        int _period_dtype_code
+
+    def __init__(self, n=1, normalize=False, month=None):
+        # Because YearEnd can be the freq for a Period, define its
+        #  _period_dtype_code at construction for performance
+        YearOffset.__init__(self, n, normalize, month)
+        self._period_dtype_code = PeriodDtypeCode.A + self.month % 12
+
 
 cdef class YearBegin(YearOffset):
     """
@@ -2099,6 +2116,14 @@ cdef class QuarterEnd(QuarterOffset):
     _prefix = "Q"
     _day_opt = "end"
 
+    cdef readonly:
+        int _period_dtype_code
+
+    def __init__(self, n=1, normalize=False, startingMonth=None):
+        # Because QuarterEnd can be the freq for a Period, define its
+        #  _period_dtype_code at construction for performance
+        QuarterOffset.__init__(self, n, normalize, startingMonth)
+        self._period_dtype_code = PeriodDtypeCode.Q_DEC + self.startingMonth % 12
 
 cdef class QuarterBegin(QuarterOffset):
     """
@@ -2148,6 +2173,7 @@ cdef class MonthEnd(MonthOffset):
     """
     DateOffset of one month end.
     """
+    _period_dtype_code = PeriodDtypeCode.M
     _prefix = "M"
     _day_opt = "end"
 
@@ -2452,6 +2478,7 @@ cdef class Week(SingleConstructorOffset):
 
     cdef readonly:
         object weekday  # int or None
+        int _period_dtype_code
 
     def __init__(self, n=1, normalize=False, weekday=None):
         BaseOffset.__init__(self, n, normalize)
@@ -2461,6 +2488,8 @@ cdef class Week(SingleConstructorOffset):
             if self.weekday < 0 or self.weekday > 6:
                 raise ValueError(f"Day must be 0<=day<=6, got {self.weekday}")
 
+            self._period_dtype_code = PeriodDtypeCode.W_SUN + (weekday + 1) % 7
+
     cpdef __setstate__(self, state):
         self.n = state.pop("n")
         self.normalize = state.pop("normalize")