[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[STINNER Victor]

STINNER Victor  added the comment:


New changeset 8d3e7eff0936926554db6162c992af5829dc8160 by Victor Stinner in 
branch 'main':
bpo-43753: _operator.is_() uses Py_Is() (GH-28641)
https://github.com/python/cpython/commit/8d3e7eff0936926554db6162c992af5829dc8160


--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[STINNER Victor]

Change by STINNER Victor :


--
pull_requests: +27009
pull_request: https://github.com/python/cpython/pull/28641

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[STINNER Victor]

STINNER Victor  added the comment:

Well, nobody came up with a better definition, so let's go with:

"Test if the x object is the y object, the same as x is y in Python."

Thanks for the feedback and reviews! Py_Is(x, y), Py_IsNone(x), Py_IsTrue(x) 
and Py_IsFalse(x) are now part of Python 3.10 C API.

--
resolution:  -> fixed
stage: patch review -> resolved
status: open -> closed

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[STINNER Victor]

STINNER Victor  added the comment:

Carl:
> Just chiming in to say that for PyPy this API would be extremely useful, 
> because PyPy's "is" is not implementable with a pointer comparison on the C 
> level (due to unboxing we need to compare integers, floats, etc by value). 
> Right now, C extension code that compares pointers is subtly broken and 
> cannot be fixed by us.

Can you come with a sentence that I can put in the documentation to explain 
that Py_Is() is written for interoperability with other Python implementations?

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[STINNER Victor]

STINNER Victor  added the comment:


New changeset 09bbebea163fe7303264cf4069c51d4d2f22fde4 by Victor Stinner in 
branch 'master':
bpo-43753: Add Py_Is() and Py_IsNone() functions (GH-25227)
https://github.com/python/cpython/commit/09bbebea163fe7303264cf4069c51d4d2f22fde4


--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[STINNER Victor]

STINNER Victor  added the comment:

> I tried applying this API on an extension, and I found the code to be 
> slightly less readable for the "is not" cases.

FYI you can try upgrade_pythoncapi.py on your project using the following PR to 
update code to use Py_IsNone/Py_IsTrue/Py_IsFalse:
https://github.com/pythoncapi/pythoncapi_compat/pull/8

For example, use "upgrade_pythoncapi.py -o Py_Is directory/" or 
"upgrade_pythoncapi.py -o Py_Is file.c".

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[Erlend Egeberg Aasland]

Erlend Egeberg Aasland  added the comment:

> I would prefer keep the C API small.

Yes, I see the value of that as well.

I tried applying this API on an extension, and I found the code to be slightly 
less readable for the "is not" cases.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[STINNER Victor]

STINNER Victor  added the comment:

> I'd also prefer a Py_IsNotNone() because it's more explicit than !Py_IsNone()

I would prefer keep the C API small. I don't think that we need to duplicate 
all functions testing for something. We provide PyTuple_Check(obj) but we don't 
provide PyTuple_NotCheck(obj) for example.

IMO !Py_IsNone(obj) makes perfectly sense in Python.

Also, "x == Py_None" is more common than "x != Py_None".

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[Erlend Egeberg Aasland]

Erlend Egeberg Aasland  added the comment:

I'd also prefer a Py_IsNotNone() because it's more explicit than !Py_IsNone(); 
the exclamation mark can be easily missed when reading/writing code, IMO.

Just my 2 cents.

--
nosy: +erlendaasland

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[STINNER Victor]

STINNER Victor  added the comment:

Mark:
> `is` is not well defined except for a small set of values, so the docs for 
> `Py_Is` would have to so vague as to be worthless, IMO.

I don't propose to change the "is" operator semantic: Py_Is(x, y) is C would 
behave *exaclty* as "x is y" in Python.

Check my implementation: the IS_OP bytecode uses directly Py_Is().

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[STINNER Victor]

STINNER Victor  added the comment:

PR 25227: I reimplemented Py_Is() as a macro and I added unit tests.

Other added functions simply call Py_Is(), example:

#define Py_IsNone(x) Py_Is(x, Py_None)

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[Raymond Hettinger]

Raymond Hettinger  added the comment:

> Just chiming in to say that for PyPy this API would be extremely useful

Thanks for that input. Given that there would be some value add, I withdraw my 
objection.

> I proposed to declare it as a "static inline" function,
> but I'm fine with a macro as well.

Let's use a macro then because inlining in guaranteed and we don't have to be 
on the lookout for failures to inline.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[Carl Friedrich Bolz-Tereick]

Carl Friedrich Bolz-Tereick  added the comment:

Just chiming in to say that for PyPy this API would be extremely useful, 
because PyPy's "is" is not implementable with a pointer comparison on the C 
level (due to unboxing we need to compare integers, floats, etc by value). 
Right now, C extension code that compares pointers is subtly broken and cannot 
be fixed by us.

--
nosy: +Carl.Friedrich.Bolz

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[STINNER Victor]

STINNER Victor  added the comment:

> Also, there is too much faith in functions marked as "inline" always being 
> inlined.

I proposed to declare it as a "static inline" function, but I'm fine with a 
macro as well. I mostly care about the API: call "Py_Is()", not really about 
the exact implementation. In practice, for now, Py_Is(x, y) will continue to be 
compiled as "x == y".

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[STINNER Victor]

STINNER Victor  added the comment:

> For any sane design of tagged pointers, `x == y` (in C) will work fine.

I wrote a proof-of-concept for using tagged pointners in CPython:
https://github.com/vstinner/cpython/pull/6

"The PR is large because of the first changes which add Py_IS_NONE(), 
Py_IS_TRUE() and Py_IS_FALSE() macros."

For backward compatibility, I added a function to get a concrete Python object 
from a tagged pointer: the True as a tagged pointer becomes (PyObject 
*)&_Py_TrueStruct concrete object. Py_IS_TRUE() has to check for both values: 
tagged pointer or &_Py_TrueStruct (don't look at my exact implementation, it's 
wrong ;-)).

In a perfect world, there would be no backward compatibility and you would 
never have to create Python objects from a tagged pointer.

The other problem is that the stable ABI exposes "Py_True" as "&_Py_TrueStruct" 
and so C extensions built with the stable ABI uses "&_Py_TrueStruct", not the 
tagged pointer.

See also bpo-39511 which discuss solutions for these problems.

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[Mark Shannon]

Mark Shannon  added the comment:

For any sane design of tagged pointers, `x == y` (in C) will work fine.

`is` is not well defined except for a small set of values, so the docs for 
`Py_Is` would have to so vague as to be worthless, IMO.

--
nosy: +Mark.Shannon

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[Mark Dickinson]

Change by Mark Dickinson :


--
nosy: +mark.dickinson

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[hai shi]

Change by hai shi :


--
nosy: +shihai1991

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[Raymond Hettinger]

Change by Raymond Hettinger :


--
nosy: +pablogsal, tim.peters

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[Raymond Hettinger]

Raymond Hettinger  added the comment:

> Right now, there is no benefit for CPython.

Please don't this until we have a clear demonstrable benefit.  As it stands 
now, this is all cost and no benefit.

Adding unnecessary abstraction layers just makes it more difficult for people 
to learn to be core devs.  Also, it will likely result in a lot of pointless 
code churn where each change carries a risk of a new bug being introduced.

The notion of pointer comparison is so fundamental to C code that it is counter 
productive to try to abstract it away.  It isn't much different than saying 
that every instance of "a + b" should be abstracted to "add(a, b)" and every 
"a[b]" should be abstracted to "index_lookup(a, b)" on the hope that maybe it 
might someday be helpful for PyPy or HPy even though they have never requested 
such a change.

>From my own point of view, these need abstractions just make it harder to tell 
>what code is actually doing.  Further, it will just lead to wasting everyone's 
>time in code reviews where the reviewer insists on the applying the new inline 
>function and there is confusion about whether two pointers are generic 
>pointers or python object pointers, each with their own comparison technique.

Also, there is too much faith in functions marked as "inline" always being 
inlined. Compilers get to make their own choices and under some circumstances 
will not inline, especially for cross module calls.  This risks taking code 
that is currently obviously fast and occasionally, invisibility slowing it down 
massively — from a step that is 1 cycle at most and is sometimes zero cost to a 
step that involves an actual function call, possibly needing to save and 
restore registers.

Lastly, the API changes aren't just for you or the standard library.  In 
effect, you're telling the entire ecosystem of C extensions that they are doing 
it wrong.  Almost certainly, some will follow this path and some won't, further 
fracturing the ecosystem.

--
nosy: +rhettinger

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[STINNER Victor]

STINNER Victor  added the comment:

Py_IS_TYPE(obj, type) was added to Python 3.9 by bpo-39573:
https://docs.python.org/dev/c-api/structures.html#c.Py_IS_TYPE

commit d905df766c367c350f20c46ccd99d4da19ed57d8
Author: Dong-hee Na 
Date:   Fri Feb 14 02:37:17 2020 +0900

bpo-39573: Add Py_IS_TYPE() function (GH-18488)

Co-Author: Neil Schemenauer 

It's currently implemented as:

static inline int _Py_IS_TYPE(const PyObject *ob, const PyTypeObject *type) {
return Py_TYPE(ob) == type;
}
#define Py_IS_TYPE(ob, type) _Py_IS_TYPE(_PyObject_CAST_CONST(ob), type)

--

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_IS(x, y) and Py_IsNone(x) functions

[STINNER Victor]

Change by STINNER Victor :


--
title: [C API] Add Py_IS(x, y) macro -> [C API] Add Py_IS(x, y) and 
Py_IsNone(x) functions

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com

[issue43753] [C API] Add Py_Is(x, y) and Py_IsNone(x) functions

[STINNER Victor]

Change by STINNER Victor :


--
title: [C API] Add Py_IS(x, y) and Py_IsNone(x) functions -> [C API] Add 
Py_Is(x, y) and Py_IsNone(x) functions

___
Python tracker 

___
___
Python-bugs-list mailing list
Unsubscribe: 
https://mail.python.org/mailman/options/python-bugs-list/archive%40mail-archive.com