ENH: Control resampling at halfyear with origin

[rwijtvliet]

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
s1 = pd.Series(1, pd.date_range('2025', freq='D', periods=700)).resample('2QS-JAN').sum()
s2 = pd.Series(1, pd.date_range('2025-04', freq='D', periods=700)).resample('2QS-JAN').sum()

# s1 expectedly has timestamps in january and july
# s1
# 2025-01-01    181
# 2025-07-01    184
# 2026-01-01    181
# 2026-07-01    154
# Freq: 2QS-JAN, dtype: int64    # NB frequency

# but s2 unexpectedly has timestamps in april and october
# s2
# 2025-04-01    183
# 2025-10-01    182
# 2026-04-01    183
# 2026-10-01    152
# Freq: 2QS-JAN, dtype: int64    # NB frequency

s1.index.freq == s2.index.freq   # True

Issue Description

It seems there is no way to force where the period boundaries are when resampling at the 2-Quarter frequency. Resampling at 2QS-APR gives the same results for s1 and s2 as those shown above.

Expected Behavior

I'd expect the index of s2 to also have timestamps on the first of January and July.

Installed Versions

INSTALLED VERSIONS ------------------ commit : 0691c5c python : 3.10.12 python-bits : 64 OS : Linux OS-release : 6.9.3-76060903-generic Version : #202405300957~1738770968~22.04~d5f7c84 SMP PREEMPT_DYNAMIC Wed F machine : x86_64 processor : x86_64 byteorder : little LC_ALL : None LANG : C.UTF-8 LOCALE : en_US.UTF-8

pandas : 2.2.3
numpy : 1.26.4
pytz : 2024.2
dateutil : 2.9.0.post0
pip : 24.3.1
Cython : None
sphinx : 7.3.7
IPython : 8.29.0
adbc-driver-postgresql: None
adbc-driver-sqlite : None
bs4 : 4.12.3
blosc : None
bottleneck : None
dataframe-api-compat : None
fastparquet : None
fsspec : None
html5lib : None
hypothesis : None
gcsfs : None
jinja2 : 3.1.4
lxml.etree : None
matplotlib : 3.9.2
numba : None
numexpr : None
odfpy : None
openpyxl : 3.1.5
pandas_gbq : None
psycopg2 : None
pymysql : None
pyarrow : None
pyreadstat : None
pytest : 8.3.3
python-calamine : None
pyxlsb : None
s3fs : None
scipy : None
sqlalchemy : None
tables : None
tabulate : 0.9.0
xarray : None
xlrd : None
xlsxwriter : None
zstandard : None
tzdata : 2024.2
qtpy : None
pyqt5 : None

[asifmohammed1]

.

[rhshadrach]

Thanks for the report, I don't think the expectation is correct. It appears to me pandas consistently resamples based on the first observation.

s = pd.Series(1, pd.date_range('2025-04-04', freq='D', periods=5))
print(s)
# 2025-04-02    1
# 2025-04-03    1
# 2025-04-04    1
# 2025-04-05    1
# 2025-04-06    1

print(s.resample('3D').sum())
# 2025-04-02    3
# 2025-04-05    2
# Freq: 3D, dtype: int64

As such, in your example, the first observation for s2 is in the April quarter, and pandas goes every 2-quarters from there on.

You can control this for certain frequencies with origin, but it has no effect for quarters. It's not clear to me if that's because we cannot support it (e.g. it's ambiguous in certain cases), do not desire to support it (e.g. complexity), or just don't yet. Further investigations are welcome!

[snitish]

@rhshadrach we currently have a QuarterBegin offset where we can specify the starting month. Would adding a new HalfYearBegin offset, with a customizable starting month, solve OP's issue? It should be relatively simple to achieve this imo.

[snitish]

Perhaps it's not a bad idea to add HalfYear offsets in general given we have Quarter and Year based offsets.

[rhshadrach]

I'm positive on this, especially if it is a simple addition.

[snitish]

take

[rwijtvliet]

Thanks guy for this fast and constructive discussion. Very happy to see this get implemented; thanks @snitish for your contribution!

[snitish]

@rwijtvliet you're welcome! However, please note - this change is still under review, so you might want to reopen the issue for tracking purposes (it'll be closed automatically after the PR is merged).