gh-119613: deprecate Py_IS_NAN/INFINITY and Py_IS_FINITE

[skirpichev]

• Issue: Deprecate Py_IS_NAN/INFINITY/FINITE? #119613

---

📚 Documentation preview 📚: https://cpython-previews--119701.org.readthedocs.build/

[erlend-aasland]

Suggested change

	use instead ``isnan``, ``isinf`` and ``isfinite`` available from ``<math.h>``
	use instead :c:func:`!isnan`, :c:func:`!isinf` and :c:func:`!isfinite` available from :file:`math.h`.

[skirpichev]

Actually, these are usually macros, e.g.:
https://en.cppreference.com/w/c/numeric/math/isinf

[erlend-aasland]

Sure; but this is about the markup :)

[vstinner]

I would prefer to soft deprecate these macros, with a link to the glossary in thr What's New and Changelog entries (:term: syntax).

[vstinner]

LGTM. IMO a soft deprecation is enough. Keeping these macros around is not a major maintenance burden.

If PEP 743 is accepted, maybe we can provide a way to remove this function with an opt-in option.

[vstinner]

Thanks for the updates @skirpichev, I enabled automerge.

@skirpichev: If you are interested, there are other macros which might be worth it to soft deprecate such as Py_MEMCPY(). I removed this macro in the limited C API version 3.11:

#if !defined(Py_LIMITED_API) || Py_LIMITED_API+0 < 0x030b0000
#  define Py_MEMCPY memcpy
#endif

[skirpichev]

@vstinner, I think automerge wont work due to canceled test (I saw similar failure in #118000).

[skirpichev]

@vstinner, we can deprecate also Py_INFINITY, Py_NAN and Py_HUGE_VAL. For the last one, probably we could also update code to use INFINITY instead. What do you think?

[vstinner]

What do you think?

You can start by avoiding these macros in Python code base and see how it goes.

Py_INFINITY and Py_NAN add a (double) cast.

[skirpichev]

Py_INFINITY and Py_NAN add a (double) cast.

Hmm, from C11 Sec 7.2: "The macro INFINITY expands to a constant expression of type float representing positive or unsigned infinity, if available [...] The macro NAN is defined if and only if the implementation supports quiet NaNs for the float type. It expands to a constant expression of type float representing a quiet NaN." (c)

As now build requirements include "Support for IEEE 754 floating point numbers and floating point Not-a-Number (NaN).", I think nowadays this type cast is useless.

Yet, anyway, replacing (with a soft deprecation) Py_HUGE_VAL by Py_INFINITY - probably does make sense. I was mentioned in #104202 (comment). One obvious pro: current using both Py_HUGE_VAL and Py_INFINITY may confuse readers.