BUG: After running `to_timestamp` on `Series` with `PeriodIndex`, frequency information is lost

[j-bennet]

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 pandas as pd
import numpy as np


def make_df(size):
    index = pd.period_range(freq="A", start="1/1/2001", periods=size)
    return pd.DataFrame({"x": np.arange(0, size)}, index=index)


df1 = make_df(2)
df2 = make_df(8)
res1 = df1.x.to_timestamp()
res2 = df2.x.to_timestamp()
print(f"{df1.x.index=}")
print(f"before: {df1.x.index.freq=}")
print(f"after: {res1.index.freq=}")
print()
print(f"{df2.x.index=}")
print(f"before: {df2.x.index.freq=}")
print(f"after: {res2.index.freq=}")

Issue Description

I'm calling to_timestamp on Series. Initially, Series has a PeriodIndex, with freq='A'. However, the Series returned from to_timestamp doesn't preserve the same freq. The behavior depends on Series size. If it's less than 3 records, the resulting Series has freq=None. If it's 3 and over, the resulting Series has freq=YearBegin.

The snippet above creates 2 dataframes, one with 2 records, and one with 8 records, and converts the series to_timestamp. It outputs index freq before and after conversion:

df1.x.index=PeriodIndex(['2001', '2002'], dtype='period[A-DEC]')
before: df1.x.index.freq=<YearEnd: month=12>
after: res1.index.freq=None

df2.x.index=PeriodIndex(['2001', '2002', '2003', '2004', '2005', '2006', '2007', '2008'], dtype='period[A-DEC]')
before: df2.x.index.freq=<YearEnd: month=12>
after: res2.index.freq=<YearBegin: month=1>

The frequency information is lost, but in a different way.

Possibly related (not the same):

• BUG: DatetimeIndex.to_period().to_timestamp() forgets freq value #38885
• BUG: Converting period index to datetime index fails when the desired frequency is at the start of a period. #48318

Expected Behavior

Since the PeriodIndex already has freq information, it should be preserved when converting to DatetimeIndex.

Installed Versions

INSTALLED VERSIONS ------------------ commit : fcb8b80 python : 3.10.8.final.0 python-bits : 64 OS : Darwin OS-release : 22.3.0 Version : Darwin Kernel Version 22.3.0: Thu Jan 5 20:50:36 PST 2023; root:xnu-8792.81.2~2/RELEASE_ARM64_T6020 machine : arm64 processor : arm byteorder : little LC_ALL : None LANG : en_US.UTF-8 LOCALE : en_US.UTF-8

pandas : 2.0.0.dev0+1448.gfcb8b809e9
numpy : 1.23.5
pytz : 2022.7.1
dateutil : 2.8.2
setuptools : 66.1.1
pip : 23.0
Cython : None
pytest : 7.2.1
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : None
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : 3.1.2
IPython : 8.9.0
pandas_datareader: None
bs4 : 4.11.2
bottleneck : None
brotli :
fastparquet : 2023.1.0
fsspec : 2023.1.0
gcsfs : None
matplotlib : None
numba : 0.56.4
numexpr : 2.8.3
odfpy : None
openpyxl : None
pandas_gbq : None
pyarrow : 11.0.0
pyreadstat : None
pyxlsb : None
s3fs : 2023.1.0
scipy : 1.10.0
snappy :
sqlalchemy : 1.4.46
tables : 3.7.0
tabulate : None
xarray : 2023.1.0
xlrd : None
zstandard : None
tzdata : None
qtpy : None
pyqt5 : None

[phofl]

Hi, thanks for your report.

This is intended since 2 values do not reliably provide frequency information. The documentation is not that easy to find though when using to_timestamp (see for example https://pandas.pydata.org/docs/reference/api/pandas.infer_freq.html#pandas-infer-freq).

Edit: Forgot to reply to the second part. to_timestamp converts to the beginning of your period per default. That's why we are changing the frequency in the second example.

Interesting enough, setting how="end" sets the freq to None, this seems like a bug.

[j-bennet]

This is intended since 2 values do not reliably provide frequency information.

If PeriodIndex can have freq with only 2 records, why can't DatetimeIndex?

to_timestamp converts to the beginning of your period per default. That's why we are changing the frequency in the second example.

But being an instance method of PeriodIndex, why does it disregard the original frequency of PeriodIndex?

[phofl]

It seems like the freq is always inferred, even if you specify it explicitly it gets overridden. Inference works only with 3 or more elements. But not really familiar enough with the conversion that happens here to judge.

We can't simply use the PeriodIndex freq, because the freq might change during the conversion

cc @MarcoGorelli any idea why a passed freq wouldn't be honoured?

[MarcoGorelli]

Thanks @j-bennet for the report

it should be preserved when converting to DatetimeIndex.

so would your expect output be

DatetimeIndex(['2001-12-31', '2002-12-31', '2003-12-31', '2004-12-31',
               '2005-12-31', '2006-12-31', '2007-12-31', '2008-12-31'],
              dtype='datetime64[ns]', freq='A-DEC')

?

Not sure that would be expected