BUG: args offset in Timestamp.__new__ causes bugs

[AlexKirko]

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

# Pandas 1.5.x
import pandas as pd

ts1 = pd.Timestamp(2023, 3, 30, 10, None, None, 30)
ts2 = pd.Timestamp(2023, 3, 30, 10, day=30)

> 2023-03-30 10:00:00.000030
> 2023-03-30 10:00:00.000030

# Pandas 2.x
ts1 = pd.Timestamp(2023, 3, 30, 10, None, 30)
ts2 = pd.Timestamp(2023, 3, 30, 10, second=30)

> 2023-03-30 10:00:00.000030
> 2023-03-30 10:00:00.000030

Issue Description

When we build Timestamp from components through positional args, the year initially ends up in ts_input, and we need to shift args one to the right, which we do with some kludges (normally by calling a constructor and passing it arguments in offset positions).

The problem is that we cannot tell the difference between the construction calls given in the example from inside the current Timestamp.__new__. This is the cause of both the bug outlined here and also of #52117 .

The two available solutions are to either decorate Timestamp, which causes a performance hit (see #52221) or to refactor Timestamp to be built from *args, **kwargs and handle the logic internally.

Expected Behavior

# Pandas 2.x
ts1 = pd.Timestamp(2023, 3, 30, 10, None, None, 30)
ts2 = pd.Timestamp(2023, 3, 30, 10, second=30)

> 2023-03-30 10:00:00.000030
> 2023-03-30 10:00:30

Installed Versions

INSTALLED VERSIONS

commit : 8c7b8a4
python : 3.8.16.final.0
python-bits : 64
OS : Windows
OS-release : 10
Version : 10.0.22621
machine : AMD64
processor : AMD64 Family 25 Model 33 Stepping 0, AuthenticAMD
byteorder : little
LC_ALL : None
LANG : None
LOCALE : English_United States.1251

pandas : 2.1.0.dev0+337.g8c7b8a4f3e
numpy : 1.23.5
pytz : 2023.2
dateutil : 2.8.2
setuptools : 67.6.0
pip : 23.0.1
Cython : 0.29.33
pytest : 7.2.2
hypothesis : 6.70.0
sphinx : 4.5.0
blosc : None
feather : None
xlsxwriter : 3.0.9
lxml.etree : 4.9.2
html5lib : 1.1
pymysql : 1.0.2
psycopg2 : 2.9.3
jinja2 : 3.1.2
IPython : 8.11.0
pandas_datareader: None
bs4 : 4.12.0
bottleneck : 1.3.7
brotli :
fastparquet : 2023.2.0
fsspec : 2023.3.0
gcsfs : 2023.3.0
matplotlib : 3.6.3
numba : 0.56.4
numexpr : 2.8.3
odfpy : None
openpyxl : 3.1.0
pandas_gbq : None
pyarrow : 11.0.0
pyreadstat : 1.2.1
pyxlsb : 1.0.10
s3fs : 2023.3.0
scipy : 1.10.1
snappy :
sqlalchemy : 2.0.7
tables : 3.7.0
tabulate : 0.9.0
xarray : 2023.1.0
xlrd : 2.0.1
zstandard : 0.19.0
tzdata : 2023.2
qtpy : None
pyqt5 : None

[AlexKirko]

As discussed in #52221, I'll be refactoring Timestamp.__new__ as a solution.

[rhshadrach]

As an alternative to what was discussed in #52221, I was wondering what we think of making ts_input keyword-only. We could still branch the logic on the type of input received for year for backwards compatibility. We would then be able to reliably get the remaining positional args even if they are specified by name.

cc @mroeschke

[mroeschke]

I think if we're considering making alterations to the constructor, I think we should make the Timestamp constructor match the datetime.datetime constructor an create from_string, from_datetime, from_unit constructors for all the other arguments it accepts.

[rhshadrach]

I took a more in depth look, it's not possible to do what I suggested; namely Timestamp(1, month=2, day=3) and Timestamp(year=1, month=2, day=3) would receive the same arguments but the result needs to differ.