Python Pandas 1.5.x takes too long to retrieve dict values, when the values are serieses using DatetimeIndex with TZ

Question

This was not an issue prior to Pandas 1.5.x.

Last version working fast as expected: 1.4.4.

The below tests demonstrate a conversion of a dict to a dataframe (df = pd.DataFrame(dict)), however, the issue exists also when performing a simple value retrieval via print(dict.values()).

The dict used for testing is a simple one where the keys are integers 0-999, and the values are pandas serieses, where each series' index is a DatetimeIndex with a 'US/Eastern' TZ and each series' values are all zeros.

The cause of the slowness is the TZ.

Tested on both Python 3.10.x and 3.11.x.

Here are the test results:

import pandas as pd
from timeit import default_timer as timer

dict = {}
index = pd.date_range(start='2020-01-01', end='2020-12-31', freq='D')

for i in range(1000):
    dict[i] = pd.Series(0, index=index)

start = timer()
df = pd.DataFrame(dict)
print(f"Pandas > 1.4.4, no TZ: {timer() - start} seconds.")

Pandas > 1.4.4, no TZ: 0.014551600441336632 seconds.

import pandas as pd
from timeit import default_timer as timer

dict = {}
index = pd.date_range(start='2020-01-01', end='2020-12-31', freq='D', tz='US/Eastern')

for i in range(1000):
    dict[i] = pd.Series(0, index=index)

start = timer()
df = pd.DataFrame(dict)
print(f"Pandas > 1.4.4, with TZ: {timer() - start} seconds.")

Pandas > 1.4.4, with TZ: 28.93334749992937 seconds.

import pandas as pd
from timeit import default_timer as timer

dict = {}
index = pd.date_range(start='2020-01-01', end='2020-12-31', freq='D')

for i in range(1000):
    dict[i] = pd.Series(0, index=index)

start = timer()
df = pd.DataFrame(dict)
print(f"Pandas 1.4.4, no TZ: {timer() - start} seconds.")

Pandas 1.4.4, no TZ: 0.023447500076144934 seconds.

import pandas as pd
from timeit import default_timer as timer

dict = {}
index = pd.date_range(start='2020-01-01', end='2020-12-31', freq='D', tz='US/Eastern')

for i in range(1000):
    dict[i] = pd.Series(0, index=index)

start = timer()
df = pd.DataFrame(dict)
print(f"Pandas 1.4.4, with TZ: {timer() - start} seconds.")

Pandas 1.4.4, with TZ: 0.06350550008937716 seconds.