BUG: frequency attribute lost if datetimeindex is localized

[rwijtvliet]

• I have checked that this issue has not already been reported.
  (Might be related to 33647, though there localization does not seem to play a role)

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

• (optional) I have confirmed this bug exists on the master branch of pandas.

---

Note: Please read this guide detailing how to provide the necessary information for us to reproduce your bug.

Code Sample, a copy-pastable example

import pandas as pd

s = pd.Series([2,5,3], index=pd.date_range('2020-01-01', periods=3, 
                                           freq='M', tz='Europe/Berlin'))
s.index #freq correctly set

DatetimeIndex(['2020-01-31 00:00:00+01:00', '2020-02-29 00:00:00+01:00',
               '2020-03-31 00:00:00+02:00'],
              dtype='datetime64[ns, Europe/Berlin]', name='ts', freq='M')

Then changing to MultiIndex:

s.index = pd.MultiIndex.from_arrays([s.index, s.index.month])
s.index.get_level_values(0)   #freq has been lost

DatetimeIndex(['2020-01-31 00:00:00+01:00', '2020-02-29 00:00:00+01:00',
               '2020-03-31 00:00:00+02:00'],
              dtype='datetime64[ns, Europe/Berlin]', name='ts', freq=None

Notice the freq=None.

Problem description

In the snipped above, the freq attribute of the DateTimeIndex is lost when a MultiIndex is created from it.

The problem seems to stem from the localization. It's equally present in DataFrames and when using set_index/tz_localize to promote a normal datetime column to index.

Expected Output

I expect the freq attribute to persist, as it does in this snippet, when I do not localize the index:

s2 = pd.Series([2,5,3], index=pd.date_range('2020-01-01', periods=3, 
                                            freq='M')) #NB no localization
s2.index = pd.MultiIndex.from_arrays([s2.index, s2.index.month]) #same as above
s2.index.get_level_values(0)

DatetimeIndex(['2020-01-31', '2020-02-29', '2020-03-31'], 
              dtype='datetime64[ns]', freq='M')

Output of pd.show_versions()

INSTALLED VERSIONS ------------------ commit : None python : 3.8.2.final.0 python-bits : 64 OS : Windows OS-release : 10 machine : AMD64 processor : Intel64 Family 6 Model 158 Stepping 10, GenuineIntel byteorder : little LC_ALL : None LANG : en LOCALE : de_DE.cp1252

pandas : 1.0.3
numpy : 1.18.1
pytz : 2019.3
dateutil : 2.8.1
pip : 20.0.2
setuptools : 46.1.3.post20200330
Cython : None
pytest : 5.4.1
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : None
IPython : 7.13.0
pandas_datareader: None
bs4 : None
bottleneck : None
fastparquet : None
gcsfs : None
lxml.etree : None
matplotlib : 3.1.3
numexpr : None
odfpy : None
openpyxl : 3.0.3
pandas_gbq : None
pyarrow : None
pytables : None
pytest : 5.4.1
pyxlsb : None
s3fs : None
scipy : 1.4.1
sqlalchemy : None
tables : None
tabulate : None
xarray : None
xlrd : 1.2.0
xlwt : None
xlsxwriter : None
numba : None

[mroeschke]

On master, it appears the second example drops frequency as well

In [6]: s2 = pd.Series([2,5,3], index=pd.date_range('2020-01-01', periods=3,
   ...:                                             freq='M')) #NB no localization
   ...: s2.index = pd.MultiIndex.from_arrays([s2.index, s2.index.month]) #same as above
   ...: s2.index.get_level_values(0)
Out[6]: DatetimeIndex(['2020-01-31', '2020-02-29', '2020-03-31'], dtype='datetime64[ns]', freq=None)

In [7]: pd.__version__
Out[7]: '1.1.0.dev0+1316.g38e1b9bc0'

cc @jbrockmendel as you've been looking at frequency stuff lately.

[jbrockmendel]

IIRC the issue here is that DatetimeIndex.factorize doesn't retain freq; it isn't obvious to me whether it should

[rwijtvliet]

Thanks guys for looking into this so quickly.

As a workaround, is there a way to re-set it, after creation of the MultiIndex?

[jbrockmendel]

As a workaround, is there a way to re-set it, after creation of the MultiIndex?

I strongly discourage this, and make no promises that it will continue to work, but

lvl = s2.index.levels[0]
lvl._data.freq = orig_freq

[giuliobeseghi]

It might be related to a bug with the tz_localize method of pd.Series:

import pandas as pd

s = pd.Series(
    [1, 2, 3, 4, 5], index=pd.date_range("2020", periods=5, freq="D")
)
print(s.index.freq)  # <Day>
print(s.tz_localize("europe/london").index.freq)  # None

[giuliobeseghi]

I just realised this is a separate bug. I'll open a new issue.

It might be related to a bug with the tz_localize method of pd.Series:

import pandas as pd

s = pd.Series(
    [1, 2, 3, 4, 5], index=pd.date_range("2020", periods=5, freq="D")
)
print(s.index.freq)  # <Day>
print(s.tz_localize("europe/london").index.freq)  # None

[giuliobeseghi]

@rwijtvliet I could recreate your problem with a naive index as well:

import pandas as pd

s = pd.Series(
    [2, 5, 3], index=pd.date_range("2020-01-01", periods=3, freq="M"),
)
s.index  # freq correctly set

s.index = pd.MultiIndex.from_arrays([s.index, s.index.month])
s.index.get_level_values(0)  # freq has been lost

Maybe the name of the issue is a bit misleading then?