Help wanted: Why do maths/radix2_fft.py test results change on Python 3.14 beta?

[cclauss]

See the required changes https://github.com/TheAlgorithms/Python/pull/12710/files#r2080117333

Repository commit

Same as #12710

Python version (python --version)

Python 3.14 beta 1

Dependencies version (pip freeze)

Same as:

• DO NOT MERGE: Test Python 3.14 #12710

Expected behavior

Doctest results do not need to change between Python 3.13 and 3.14 beta.

• DO NOT MERGE: Test Python 3.14 #12710

Actual behavior

See changes required in:

• DO NOT MERGE: Test Python 3.14 #12710

[MaximSmolskiy]

@cclauss In Python 3.14 Release Notes (https://docs.python.org/3.14/whatsnew/3.14.html) there are mentions of changes related to complex numbers:

• Implement mixed-mode arithmetic rules combining real and complex numbers as specified by C standards since C99. (Contributed by Sergey B Kirpichev in gh-69639.)

This may be the reason for changes in the doctests.

[cclauss]

@skirpichev

[skirpichev]

JFR, above pr was merged not before latest beta. Did you test alphas?

[cclauss]

https://github.com/TheAlgorithms/Python/actions/runs/14991034991/job/42114180325#step:5:20 This is beta 1.

Or is uv using a different Py3.14?

[skirpichev]

My question was did you test 3.14 alphas?

[skirpichev]

Anyway, so far I can guess this might be related to:

        inverce_c = [round(x[0].real, 8) + round(x[0].imag, 8) * 1j for x in inverce_c]

Probably, you want rather complex(round(x[0].real, 8), round(x[0].imag, 8)). But with above pattern you can get different answers on 3.14 and 3.13. Example:

Python 3.14.0a7 (tags/v3.14.0a7:29af6cee02, Apr  9 2025, 12:48:04) [GCC 12.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> 1 + (-0.0)*1j
(1-0j)
Python 3.13.3 (tags/v3.13.3:6280bb5478, Apr  9 2025, 10:07:49) [GCC 12.2.0] on linux
Type "help", "copyright", "credits" or "license" for more information.
>>> 1 + (-0.0)*1j
(1+0j)

Obviously, the last answer is incorrect as a replacement of complex(1, -0.0).