natbusa · January 20, 2022 03:09
diff --git a/compare_offsets.py b/compare_offsets.py
 # Comparing pandas DateOffset strings directly one another is not available 
 # as a method of the DateOffset class in python. The reason is straightforward:
 # offsets are "timestamp context" dependent. 

 # For example 1M can be either 28D, 29D, 30D, or 31D so there is no way to compare
 # *exactly * DateOffset strings with each other out of context. 

 # To alleviate this, here below you can find a reference implementation 
 # for an approximate comparison assuming months of 30 days, and years of 365 days.

 def alias_to_days(x):
    return {
        "B":     1,
        "D":     1,
        "W":     7,
        "M":    30,
        "SM":   15,
        "BM":   30,
        "MS":   30,
        "SMS":  15,
        "BMS":  30,
        "Q":    90,
        "BQ":   90,
        "QS":   90,
        "BQS":  90,
        "A":   365,
        "Y":   365,
        "BA":  365,
        "BY":  365,
        "AS":  365,
        "YS":  365,
        "BAS": 365,
        "BYS": 365,
    }.get(x, None)

 def alias_to_seconds(x):
    return {
        "BH":  3600,
        "H":   3600,
        "T":     60,
        "min":   60,
        "S":      1,
    }.get(x)

 def alias_to_int(x):
    n= to_days(x)
    if n is not None:
        return n * 86400
    
    n= to_seconds(x)
    if n is not None:
        return n
    
    raise ValueError(f'{x} is not a valid/supported string offset')


 import re
 def offset_to_int(x):
    i = 0
    
    first=r"^(\d*)([a-zA-Z]+)(-[a-zA-Z]+)?(.*)$"
    
    g = re.match(first,x).groups()
    i += (int(g[0]) or 1)* alias_to_int(g[1])
    
    # handle concatenated datetime offset strings
    if g[3]:
        block=r"(\d+)([a-zA-Z]+)(-[a-zA-Z]+)?"
        for e in re.findall(block,g[3]):
            i += (int(e[0]) or 1)* alias_to_int(e[1])
    return i

 def compare_offsets(a, b):
    if a==b:
        return 0
    
    a = offset_to_int(a)
    b = offset_to_int(b)
    
    if a==b:
        return 0
    elif  a>b:
        return 1
    else:
        return -1
diff --git a/examples.txt b/examples.txt
 >> compare_offsets('72H30T', '3D')
 1

 >> compare_offsets('72H', '3D')
 0

 >> compare_offsets('2B12BH', '3D')
 -1

 offset_to_int('2QS-JUN12W30T-MON69S1BH')
 22815069
	# Comparing pandas DateOffset strings directly one another is not available
	# as a method of the DateOffset class in python. The reason is straightforward:
	# offsets are "timestamp context" dependent.

	# For example 1M can be either 28D, 29D, 30D, or 31D so there is no way to compare
	# exactly DateOffset strings with each other out of context.

	# To alleviate this, here below you can find a reference implementation
	# for an approximate comparison assuming months of 30 days, and years of 365 days.

	def alias_to_days(x):
	return {
	"B": 1,
	"D": 1,
	"W": 7,
	"M": 30,
	"SM": 15,
	"BM": 30,
	"MS": 30,
	"SMS": 15,
	"BMS": 30,
	"Q": 90,
	"BQ": 90,
	"QS": 90,
	"BQS": 90,
	"A": 365,
	"Y": 365,
	"BA": 365,
	"BY": 365,
	"AS": 365,
	"YS": 365,
	"BAS": 365,
	"BYS": 365,
	}.get(x, None)

	def alias_to_seconds(x):
	return {
	"BH": 3600,
	"H": 3600,
	"T": 60,
	"min": 60,
	"S": 1,
	}.get(x)

	def alias_to_int(x):
	n= to_days(x)
	if n is not None:
	return n * 86400

	n= to_seconds(x)
	if n is not None:
	return n

	raise ValueError(f'{x} is not a valid/supported string offset')


	import re
	def offset_to_int(x):
	i = 0

	first=r"^(\d)([a-zA-Z]+)(-[a-zA-Z]+)?(.)$"

	g = re.match(first,x).groups()
	i += (int(g[0]) or 1)* alias_to_int(g[1])

	# handle concatenated datetime offset strings
	if g[3]:
	block=r"(\d+)([a-zA-Z]+)(-[a-zA-Z]+)?"
	for e in re.findall(block,g[3]):
	i += (int(e[0]) or 1)* alias_to_int(e[1])
	return i

	def compare_offsets(a, b):
	if a==b:
	return 0

	a = offset_to_int(a)
	b = offset_to_int(b)

	if a==b:
	return 0
	elif a>b:
	return 1
	else:
	return -1
	>> compare_offsets('72H30T', '3D')
	1

	>> compare_offsets('72H', '3D')
	0

	>> compare_offsets('2B12BH', '3D')
	-1

	offset_to_int('2QS-JUN12W30T-MON69S1BH')
	22815069