BUG: infer_freq has stateful behavior

[mdeceglie]

Pandas version checks

• I have checked that this issue has not already been reported.

• I have confirmed this bug exists on the latest version of pandas.

• I have confirmed this bug exists on the main branch of pandas.

Reproducible Example

import datetime
import pandas as pd
tz = 'America/Santiago'
start_date = datetime.datetime(2018, 8, 10, 0, 0, 0)
end_date = datetime.datetime(2018, 8, 14, 23, 0, 0)
freq = 'H'
times = pd.date_range(start=start_date, end=end_date, freq=freq)
times = times.tz_localize(tz=tz, ambiguous='infer',
                          nonexistent='shift_forward')
print(pd.infer_freq(times[:10]))
pd.infer_freq(times)
print(pd.infer_freq(times[:10]))

Issue Description

Initially, infer_freq on the first 10 items of the index returns H, after attempting it on the full index, it returns None on the first 10 items of the index. Confirmed expected behavior in version 2.0.3.

Expected Behavior

Return H in both instances of pd.infer_freq(times[:10]) in the example.

Installed Versions

INSTALLED VERSIONS

commit : a60ad39
python : 3.10.13.final.0
python-bits : 64
OS : Darwin
OS-release : 21.6.0
Version : Darwin Kernel Version 21.6.0: Wed Oct 4 23:56:02 PDT 2023; root:xnu-8020.240.18.704.15~1/RELEASE_ARM64_T6000
machine : x86_64
processor : i386
byteorder : little
LC_ALL : None
LANG : en_US.UTF-8
LOCALE : en_US.UTF-8

pandas : 2.1.2
numpy : 1.26.0
pytz : 2023.3.post1
dateutil : 2.8.2
setuptools : 68.0.0
pip : 23.3
Cython : None
pytest : 7.4.3
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : 4.9.3
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : 3.1.2
IPython : 8.15.0
pandas_datareader : None
bs4 : 4.12.2
bottleneck : None
dataframe-api-compat: None
fastparquet : None
fsspec : None
gcsfs : None
matplotlib : 3.8.0
numba : None
numexpr : None
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : None
pyreadstat : None
pyxlsb : None
s3fs : None
scipy : 1.11.3
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : None
zstandard : None
tzdata : 2023.3
qtpy : 2.2.0
pyqt5 : None

[kandersolar]

Perhaps this was introduced in #51738? A call to times._engine.clear_mapping() seems to fix things:

times = pd.date_range(start="2018-08-11 20:00", end="2018-08-12 04:00", freq="H")
times = times.tz_localize(tz="America/Santiago", ambiguous='infer',
                          nonexistent='shift_forward')

print(pd.infer_freq(times[:3]))  # H
pd.infer_freq(times)
print(pd.infer_freq(times[:3]))  # None
times._engine.clear_mapping()
print(pd.infer_freq(times[:3]))  # H

Here is a related example:

times = pd.date_range(start="2018-08-11 20:00", end="2018-08-12 04:00", freq="H")
times = times.tz_localize(tz="America/Santiago", ambiguous='infer', nonexistent='shift_forward')

print(times[:3]._is_unique)  # True
times._is_unique
print(times[:3]._is_unique)  # False
times._engine.clear_mapping()
print(times[:3]._is_unique)  # True

This times DatetimeIndex contains equivalent/duplicate times. The tested slice does not, but incorrectly inherits the cached determination of non-uniqueness from its parent. Perhaps a suitable fix is to make slices not inherit unique and need_unique_check from the parent index?

[kandersolar]

A slightly more minimal reproducer:

# last datetime is a duplicate
times = pd.to_datetime(['2019-01-01', '2019-01-02', '2019-01-03', '2019-01-03'])

print(pd.infer_freq(times[:3]))  # D
pd.infer_freq(times)
print(pd.infer_freq(times[:3]))  # None
times._engine.clear_mapping()
print(pd.infer_freq(times[:3]))  # D