How to create a rolling window of fixed length in pandas?

Question

Given a timeseries, how do I create a rolling window of some interval such that it starts with that same interval, instead of expanding from size 1. As seen here:

import pandas as pd
from datetime import timedelta

# 'per' x 1 minute (1T) intervals 
per = 10
d = pd.DataFrame(
    {'a': list(range(per))}, 
    index=pd.date_range('2021-05-01T0000', freq='1T', periods=per))

# create a rolling 5 minute window and get its length
w = d.rolling(timedelta(minutes=5))
for wi in w:
    print(len(wi))

# Output (window lengths):
# Window starts with length 1 and iterates until it expands
# to desired size:
#1
#2
#3
#4
#5  < I only want windows starting here
#5
#5
#5
#5

How to instead start the window with the specified window size?

Answer 1

Maybe a "workaround": you can use itertools.dropwhile to filter the windows with length < 5:

from itertools import dropwhile

w = dropwhile(lambda w: len(w) < 5, d.rolling(timedelta(minutes=5)))
for wi in w:
    print(len(wi))

Prints:

5
5
5
5
5
5