BUG: TypeError("type Timedelta doesn't define __round__ method")

[JustAnotherVeryNormalDeveloper]

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

class CustomTimedelta(pd.Timedelta):  
    def __round__(self, n=None):  
        seconds = self.total_seconds()
        rounded_seconds = round(seconds, n) if n else round(seconds)
        return pd.to_timedelta(rounded_seconds, unit='s') 

# Uncomment this to solve the problem
# pd.Timedelta = CustomTimedelta

# Example usage  
td = pd.Timedelta(days=1, hours=2, minutes=34, seconds=56.78)
print(round(td))  # prints '1 days 02:35:00'

Issue Description

I try to round the value, actually I want to round for the seconds. I was surprised to see that there are no default round method ?
Best use of round would be to precise the type of round (day/seconds/...) but I think it should have a default behavior.
In a pythonic way.
And I think seconds should be the default.

Expected Behavior

It should have a default behavior, in a pythonic way. I propose seconds for the default.

Installed Versions

INSTALLED VERSIONS

commit : fd3f571
python : 3.9.19.final.0
python-bits : 64
OS : Linux
OS-release : 4.4.0-19041-Microsoft
Version : #3996-Microsoft Thu Jan 18 16:36:00 PST 2024
machine : x86_64
processor : x86_64
byteorder : little
LC_ALL : None
LANG : fr_FR.UTF-8
LOCALE : fr_FR.UTF-8

pandas : 2.2.0
numpy : 1.26.3
pytz : 2024.1
dateutil : 2.9.0.post0
setuptools : 69.2.0
pip : 24.0
Cython : None
pytest : None
hypothesis : None
sphinx : None
blosc : None
feather : None
xlsxwriter : None
lxml.etree : 5.1.0
html5lib : None
pymysql : None
psycopg2 : None
jinja2 : 3.1.3
IPython : 8.18.1
pandas_datareader : None
adbc-driver-postgresql: None
adbc-driver-sqlite : None
bs4 : 4.12.3
bottleneck : None
dataframe-api-compat : None
fastparquet : None
fsspec : None
gcsfs : None
matplotlib : 3.8.2
numba : None
numexpr : None
odfpy : None
openpyxl : 3.1.2
pandas_gbq : None
pyarrow : None
pyreadstat : None
python-calamine : None
pyxlsb : None
s3fs : None
scipy : 1.12.0
sqlalchemy : 2.0.28
tables : None
tabulate : None
xarray : None
xlrd : None
zstandard : None
tzdata : 2024.1
qtpy : None
pyqt5 : None

[Aloqeely]

Does the documentation say that Timedelta can be rounded? If not, I think this issue should be labeled as a feature request rather than a bug report, with a clear title that describes the desired feature.

[JustAnotherVeryNormalDeveloper]

yes it can @Aloqeely :)
(https://pandas.pydata.org/docs/reference/api/pandas.Timedelta.round.html)
The thing is that it doesn't have a default value yet, you have to precise the parameter freq.
So you are right to say it's more about a feature rater than a bug.

Yet, I find it weird to crash on a simple round() when the library think about doing serie.round() !

[rhshadrach]

When reading code, I do not think it is clear what round(td, n=5) will do, as opposed to td.round(freq="5s"). As such, I do not think we should support Python's round.

[Aloqeely]

I believe the suggestion was to have a default behavior for Python's round on timedeltas such that round(td) works without any extra args.
And seconds being the default freq makes some sense as seconds are the fractional part of the timedelta
(I.E. td = 15 days 07:22:56.782037999 --> round(td) = 15 days 07:22:57)

[JustAnotherVeryNormalDeveloper]

Perfect reformulation @Aloqeely :)
The question is still "should we do it ?" or "how to do it ?" ?

For the "should we do it ?" I would say seconds is a logical default choice. And we like default behavior in python even if we all prefer specific precision, it's always appreciable to make code work in a really short way.

For the 'how to do it ?" I would just Override the round method as I did on my custom class (and so I won't have to do it in my project and the future projects but the library would handle it). But it's maybe too simple and I miss side effects ?

Should I propose a pull request ?

[jbrockmendel]

Based on discussion in #58081 can we close this as no-action?

[mroeschke]

Yeah thanks for the suggestion but closing