BUG: 'infer_freq' does not work with tz != "UTC"

[sdementen]

The issue has been reported in #8772 which has been closed asking for an example (#8772 (comment)).

Here is the example

import pandas

for tz in [None, "UTC", "CET"]:
    index = pandas.date_range("2018", "2019", tz=tz, closed="left", freq='H')
    print(f"tz = {tz}, index freq = {index.freq}, inferred index freq = {pandas.infer_freq(index)}")

Problem description

pandas.infer_freq can infer the frequency when the index has no tz or tz=="UTC". But for tz=="CET", it returns None.

This is due to a special handling in https://github.com/pandas-dev/pandas/blob/master/pandas/tseries/frequencies.py#L198 to cater for timezone DST in frequencies beyond the hour (day, month,...). However for regular hourly frequencies, the delta which is constant when considering real time/UTC will now show multiple deltas (0 for the DST in october, 2H for the DST in march, 1H for the rest).
Probably needs two paths, one for business frequencies (days, months, ...) and one for hour and below frequencies. Not sure about frequencies between H and D (like 6H) but probably to be considered as business frequencies.

Expected Output

# output
# tz = None, index freq = <Hour>, inferred index freq = H
# tz = UTC, index freq = <Hour>, inferred index freq = H
# tz = CET, index freq = <Hour>, inferred index freq = None

Output of pd.show_versions()

INSTALLED VERSIONS

commit : b5958ee
python : 3.8.5.final.0
python-bits : 64
OS : Windows
OS-release : 10
Version : 10.0.18362
machine : AMD64
processor : Intel64 Family 6 Model 142 Stepping 10, GenuineIntel
byteorder : little
LC_ALL : None
LANG : None
LOCALE : French_Belgium.1252

pandas : 1.1.5
numpy : 1.19.4
pytz : 2020.4
dateutil : 2.8.1
pip : 20.2.4
setuptools : 50.3.2
Cython : 0.29.17
pytest : 6.1.2
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : 4.6.2
html5lib : None
pymysql : None
psycopg2 : 2.8.6 (dt dec pq3 ext lo64)
jinja2 : 2.11.2
IPython : 7.19.0
pandas_datareader: None
bs4 : 4.9.3
bottleneck : None
fsspec : 0.8.4
fastparquet : 0.4.2
gcsfs : None
matplotlib : 3.3.4
numexpr : 2.7.2
odfpy : None
openpyxl : 3.0.5
pandas_gbq : None
pyarrow : 2.0.0
pytables : None
pyxlsb : None
s3fs : None
scipy : 1.5.4
sqlalchemy : None
tables : 3.6.1
tabulate : None
xarray : None
xlrd : 1.2.0
xlwt : None
numba : 0.52.0

[sdementen]

my current workaround to bypass the conversion

index = pandas.date_range("2018", "2019", tz="CET", closed="left", freq='H')
freq = pandas.infer_freq(idx.tz_convert("UTC"))

[MarcoGorelli]

Thanks @sdementen - you have an old version of pandas (1.1.5), but I can confirm this reproduces on master

[sdementen]

I am working on a simple fix/PR