Breaking change Python 3.13/3/11

[sjoerd0301]

Bug report

Bug description:

      pandas/_libs/tslibs/conversion.cpython-313-x86_64-linux-gnu.so.p/pandas/_libs/tslibs/conversion.pyx.c: In function ‘__Pyx_PyInt_As_npy_int64’:
      pandas/_libs/tslibs/conversion.cpython-313-x86_64-linux-gnu.so.p/pandas/_libs/tslibs/conversion.pyx.c:31168:27: error: too few arguments to function ‘_PyLong_AsByteArray’

I cannot easily install an older python version, as this is dictated by my OS. I cannot install another version of CPython, because this is dictated by the project I'm trying to run.

CPython versions tested on:

3.13

Operating systems tested on:

Linux

[StanFromIreland]

This is a bug in pandas and not cpython from what I can tell.

Edit: It is, there are several similar reports in their issue tracker: https://github.com/pandas-dev/pandas/issues

[sjoerd0301]

Every project which uses cpython _PyLong_AsByteArray has this issue.
The method signature has been changed in a minor version release.

Edit: there are no issues on the pandas tracker; https://github.com/pandas-dev/pandas/issues?q=is%3Aissue%20_PyLong_AsByteArray

[StanFromIreland]

Pandas will update it when they support 3.13. The currently only support up to 3.12.

[hoodmane]

Pandas will update it when they support 3.13. The currently only support up to 3.12.

Thankfully, this is not correct:
https://pypi.org/project/pandas/#pandas-2.2.3-cp313-cp313-manylinux_2_17_x86_64.manylinux2014_x86_64.whl

[StanFromIreland]

@hoodmane

As per their docs:

Officially Python 3.9, 3.10, 3.11 and 3.12.

https://pandas.pydata.org/docs/getting_started/install.html#python-version-support

[hoodmane]

Well they publish 3.13 wheels, run the unit tests against 3.13, and include 'Programming Language :: Python :: 3.13' in their classifiers. Is it possible they just forgot to update the text of install.html? If not, I would think they should call this out more obviously. I guess I'll ask on their issue tracker.

[da-woods]

Every project which uses cpython _PyLong_AsByteArray has this issue.

It's internal. It's allowed to change. That's what the underscore at the start means.

---

However: this'll be because you're trying to build Pandas with an older Cython version. This has been fixed in Cython for a while.

[hoodmane]

Okay I opened pandas-dev/pandas#61139 asking about the specific Python 3.13 support question.

[adnit]

I do have the issue too

[picnixz]

Even if pandas should support 3.13, the stability of private functions should not be relied upon as private functions are not subject to backwards compatibility policies, even though we try to avoid breaking them when possible. Note that changing it back to what it was would be breaking for 3.13+, which is worse IMO as <3.12 only accept security fixes.

For now, I'll close this on CPython's side as not an issue.