Consider adding public PyLong_FlipSign() function

[skirpichev]

Feature or enhancement

Proposal:

Currently there is only a slow way (i.e. PyNumber_Negative()) to invert the sign of a big integer. Direct manipulation of the sign value is available only via private function, e.g. _PyLong_SetSignAndDigitCount():
https://github.com/aleaxit/gmpy/blob/1c0d4f38d4f1093df0e9460d82e4331d3d3946b6/src/gmpy2_convert_gmp.c#L153

I propose this interface:

int PyLong_FlipSign(PyObject *obj);

For small integers it will fallback to PyLong_FromLong() (like _PyLong_Negate() does), for the rest - just _PyLong_FlipSign() to change sign in place.

Alternatively, we could just expose a direct wrapper of _PyLong_FlipSign() and warn users to not use this function for small ints (that doesn't make sense for me anyway).

In both variants, on success - flip the sign of obj and return 0. On failure - return -1 with exception set. Shouldn't fail on PyLongObject or it's subtype.

I'll work on a patch if this proposal does make sense.

Has this already been discussed elsewhere?

This is a minor feature, which does not need previous discussion elsewhere

Links to previous discussion of this feature:

No response

Linked PRs

• gh-120446: add PyLong_FlipSign() public function #120489

[serhiy-storchaka]

I planned to propose adding PyLong_NegativeInPlace() and PyLong_AbsoluteInPlace().

[skirpichev]

Hmm, it seems that my first variant can't be implemented with same signature as the second. Rather it should be (like _PyLong_Negate()):

int PyLong_FlipSign(PyObject **obj);

[skirpichev]

adding PyLong_NegativeInPlace() and PyLong_AbsoluteInPlace().

I think this naming better fits case, when there will be also other PyNumber_*-like functions. But I'm not sure that we have practical need for that.

FYI, PR: #120489 (as a draft; IIUC, this should be discussed in C-API WG)

[skirpichev]

PR closed in favor of #121339.

Now I don't have a strong motivation for this function. Kept open as @serhiy-storchaka planned a similar proposal.

[vstinner]

Python int objects are immutable. I dislike APIs which seem to treat them as mutable objects.

What's the status of this issue? PEP 757 has been implementation in Python 3.14. Can we close this issue?

[encukou]

Proposal should go to the C API WG (ideally after public discussion)