Mailman 3 Poll: Py_REPLACE/Py_ASSIGN/etc - Python-Dev

newer
Call for Python Developers for our...

Poll: Py_REPLACE/Py_ASSIGN/etc

Serhiy Storchaka

26 Feb 2014 26 Feb '14

1:40 a.m.

There were several suggestions for naming new macros which replace old value with new value and then (x)decref old value. #define Py_XXX(ptr, value) \ { \ PyObject *__tmp__ = ptr; \ ptr = new_value; \ Py_DECREF(__tmp__); \ } Listed in order of receipt: 1. Py_(X)SETREF. 2. Py_(X)ASSIGN. 3. Py_(X)DECREC_REPLACE. 4. Py_REF_ASSIGN (Py_XREF_ASSIGN? Py_REF_XASSIGN?). 5. Py_(X)REPLACE. 6. Py_(X)STORE 7. Py_CLEAR_AND_SET. 8. Py_SET_AND_(X)DECREF. 9. Py_(X)DECREF_AND_ASSIGN. 10. Py_ASSIGN_AND_(X)DECREF. 11. Other... Let's choose the least confusing names. See discussions at: http://bugs.python.org/issue3081 http://bugs.python.org/issue16447 http://bugs.python.org/issue20440 http://comments.gmane.org/gmane.comp.python.devel/145346

Show replies by thread

Ryan Gonzalez

26 Feb 26 Feb

8:09 a.m.

I like Py_DECREF_REPLACE. It gives the impression that it decrefs the original and replaces it. On Wed, Feb 26, 2014 at 3:40 AM, Serhiy Storchaka wrote:

...

There were several suggestions for naming new macros which replace old value with new value and then (x)decref old value.

#define Py_XXX(ptr, value) \ { \ PyObject *__tmp__ = ptr; \ ptr = new_value; \ Py_DECREF(__tmp__); \ }

Listed in order of receipt:

1. Py_(X)SETREF. 2. Py_(X)ASSIGN. 3. Py_(X)DECREC_REPLACE. 4. Py_REF_ASSIGN (Py_XREF_ASSIGN? Py_REF_XASSIGN?). 5. Py_(X)REPLACE. 6. Py_(X)STORE 7. Py_CLEAR_AND_SET. 8. Py_SET_AND_(X)DECREF. 9. Py_(X)DECREF_AND_ASSIGN. 10. Py_ASSIGN_AND_(X)DECREF. 11. Other...

Let's choose the least confusing names.

See discussions at:

http://bugs.python.org/issue3081 http://bugs.python.org/issue16447 http://bugs.python.org/issue20440 http://comments.gmane.org/gmane.comp.python.devel/145346

_______________________________________________ Python-Dev mailing list Python-Dev@python.org https://mail.python.org/mailman/listinfo/python-dev Unsubscribe: https://mail.python.org/mailman/options/python-dev/ rymg19%40gmail.com

-- Ryan If anybody ever asks me why I prefer C++ to C, my answer will be simple: "It's becauseslejfp23(@#Q*(E*EIdc-SEGFAULT. Wait, I don't think that was nul-terminated."

Antoine Pitrou

10:26 a.m.

On Wed, 26 Feb 2014 11:40:01 +0200 Serhiy Storchaka wrote:

...

There were several suggestions for naming new macros which replace old value with new value and then (x)decref old value.

#define Py_XXX(ptr, value) \ { \ PyObject *__tmp__ = ptr; \ ptr = new_value; \ Py_DECREF(__tmp__); \ }

...

1. Py_(X)SETREF.

My vote is on this one. I'm also -1 on any name which doesn't have "REF" in it; the name should clearly suggest that it's a refcounting operation. Regards Antoine.

Nick Coghlan

4:11 p.m.

On 27 Feb 2014 04:28, "Antoine Pitrou" wrote:

...

On Wed, 26 Feb 2014 11:40:01 +0200 Serhiy Storchaka wrote:

...
There were several suggestions for naming new macros which replace old value with new value and then (x)decref old value.

#define Py_XXX(ptr, value) \ { \ PyObject *__tmp__ = ptr; \ ptr = new_value; \ Py_DECREF(__tmp__); \ }

...
1. Py_(X)SETREF.

My vote is on this one. I'm also -1 on any name which doesn't have "REF" in it; the name should clearly suggest that it's a refcounting operation.

Yeah, I think SETREF is my favourite as well (even though some of the later suggestions were mine). Cheers, Nick.

...

Regards

Antoine.

_______________________________________________ Python-Dev mailing list Python-Dev@python.org https://mail.python.org/mailman/listinfo/python-dev Unsubscribe:

https://mail.python.org/mailman/options/python-dev/ncoghlan%40gmail.com

Georg Brandl

11:13 p.m.

Am 26.02.2014 17:09, schrieb Ryan Gonzalez:

...

I like Py_DECREF_REPLACE. It gives the impression that it decrefs the original and replaces it.

Agreed, most other suggestions are not really explicit enough. Georg

Kristján Valur Jónsson

27 Feb 27 Feb

5:10 a.m.

I agree with NICK. having REF in it is a good idea. So, I'm +1 on setref. Having long explicit macros with exact semantics in the name is a bad one. so I'm -1 on any Py_DECREF_AND_REPLACE or similar daschhunds. Also, is there any real requirement for having separate non-X versions of these? The Xs constitue a permutation explosion, particularly if you want then also versions that INCREF the source :) K From: Python-Dev [mailto:python-dev-bounces+kristjan=ccpgames.com@python.org] On Behalf Of Nick Coghlan Sent: 27. febrúar 2014 00:12 To: Antoine Pitrou Cc: python-dev@python.org Subject: Re: [Python-Dev] Poll: Py_REPLACE/Py_ASSIGN/etc On 27 Feb 2014 04:28, "Antoine Pitrou" mailto:solipsis@pitrou.net> wrote:

...

On Wed, 26 Feb 2014 11:40:01 +0200 Serhiy Storchaka mailto:storchaka@gmail.com> wrote:

...
There were several suggestions for naming new macros which replace old value with new value and then (x)decref old value.

#define Py_XXX(ptr, value) \ { \ PyObject *__tmp__ = ptr; \ ptr = new_value; \ Py_DECREF(__tmp__); \ }

...
1. Py_(X)SETREF.

My vote is on this one. I'm also -1 on any name which doesn't have "REF" in it; the name should clearly suggest that it's a refcounting operation.

Yeah, I think SETREF is my favourite as well (even though some of the later suggestions were mine). Cheers, Nick.

...

Regards

Antoine.

_______________________________________________ Python-Dev mailing list Python-Dev@python.orgmailto:Python-Dev@python.org https://mail.python.org/mailman/listinfo/python-dev Unsubscribe: https://mail.python.org/mailman/options/python-dev/ncoghlan%40gmail.com

Larry Hastings

28 Feb 28 Feb

12:56 a.m.

On 02/26/2014 11:13 PM, Georg Brandl wrote:

...

Am 26.02.2014 17:09, schrieb Ryan Gonzalez:

...
I like Py_DECREF_REPLACE. It gives the impression that it decrefs the original and replaces it. Agreed, most other suggestions are not really explicit enough.

+1 from me too. When I saw Py_SETREF I thought, oh, it sets the thing and increfs it. FWIW this vote is just on the name. I haven't stared at the whole Py_REPLACE idea enough to have an opinion about whether or not to use it. But if we use it I'm +1 on Py_DECREF_REPLACE. //arry/

Nick Coghlan

4:27 a.m.

On 28 Feb 2014 19:05, "Larry Hastings" wrote:

...

On 02/26/2014 11:13 PM, Georg Brandl wrote:

...
Am 26.02.2014 17:09, schrieb Ryan Gonzalez:

...
I like Py_DECREF_REPLACE. It gives the impression that it decrefs the

original

...

...
...
and replaces it.

Agreed, most other suggestions are not really explicit enough.

+1 from me too. When I saw Py_SETREF I thought, oh, it sets the thing and increfs it.

FWIW this vote is just on the name. I haven't stared at the whole Py_REPLACE idea enough to have an opinion about whether or not to use it. But if we use it I'm +1 on Py_DECREF_REPLACE.

For additional context, the idea itself is necessary for the same reason Py_CLEAR was added: to help ensure that an object's state is never pointing at another object that is in the process of being deleted. The difference is that Py_CLEAR only allows setting the pointer to NULL, while the point of the new macro is to set it to an arbitrary existing point. There is no implicit incref as that isn't needed for correctness (you can do the incref before the pointer replacement, and often the reference count will already be correct without an explicit incref anyway). With the new macro in place, the existing Py_CLEAR(x) macro would be equivalent to Py_SETREF(x, NULL). Originally I was also concerned about the "how will people know there's no implicit incref?", but I've since become satisfied with the fact that the precedent set by the reference stealing SET_ITEM macros is strong enough to justify the shorter name. Cheers, Nick.

Kristján Valur Jónsson

5:58 a.m.

+1 Also, for the equivalence to hold there is no separate Py_XSETREF, the X behaviour is implied, which I favour. Enough of this X-proliferation already! But also see the discussion on inlines. It would be great to make this an inline rather than a macro. K From: Python-Dev [mailto:python-dev-bounces+kristjan=ccpgames.com@python.org] On Behalf Of Nick Coghlan Sent: 28. febrúar 2014 12:27 To: Larry Hastings Cc: python-dev@python.org Subject: Re: [Python-Dev] Poll: Py_REPLACE/Py_ASSIGN/etc For additional context, the idea itself is necessary for the same reason Py_CLEAR was added: to help ensure that an object's state is never pointing at another object that is in the process of being deleted. The difference is that Py_CLEAR only allows setting the pointer to NULL, while the point of the new macro is to set it to an arbitrary existing point. There is no implicit incref as that isn't needed for correctness (you can do the incref before the pointer replacement, and often the reference count will already be correct without an explicit incref anyway). With the new macro in place, the existing Py_CLEAR(x) macro would be equivalent to Py_SETREF(x, NULL). Originally I was also concerned about the "how will people know there's no implicit incref?", but I've since become satisfied with the fact that the precedent set by the reference stealing SET_ITEM macros is strong enough to justify the shorter name. Cheers, Nick.

Barry Warsaw

7:19 a.m.

On Feb 28, 2014, at 10:27 PM, Nick Coghlan wrote:

...

With the new macro in place, the existing Py_CLEAR(x) macro would be equivalent to Py_SETREF(x, NULL).

Originally I was also concerned about the "how will people know there's no implicit incref?", but I've since become satisfied with the fact that the precedent set by the reference stealing SET_ITEM macros is strong enough to justify the shorter name.

I haven't had time to follow this discussion at all, but for a macro to be called Py_SETREF and *not* increment the reference counter seems at best confusing. Despite my hesitation to paint a bike shed I haven't had time to inspect, something more akin to Py_SET_POINTER seems more appropriate (i.e. don't put "REF" in the name if it isn't playing refcounting games). -Barry

Nick Coghlan

2:15 p.m.

On 1 Mar 2014 01:22, "Barry Warsaw" wrote:

...

On Feb 28, 2014, at 10:27 PM, Nick Coghlan wrote:

...
With the new macro in place, the existing Py_CLEAR(x) macro would be equivalent to Py_SETREF(x, NULL).

Originally I was also concerned about the "how will people know there's

...

...
implicit incref?", but I've since become satisfied with the fact that the precedent set by the reference stealing SET_ITEM macros is strong enough to justify the shorter name.

I haven't had time to follow this discussion at all, but for a macro to be called Py_SETREF and *not* increment the reference counter seems at best confusing. Despite my hesitation to paint a bike shed I haven't had time to inspect, something more akin to Py_SET_POINTER seems more appropriate (i.e. don't put "REF" in the name if it isn't playing refcounting games).

It *is* playing refcounting games - it's decrefing the existing target while stealing a reference to the new target, just like the existing SET_ITEM macros and somewhat like Py_CLEAR (although in that case, it's more obvious that we will never incref NULL). The whole point of this macro is to take an *existing* reference and safely *overwrite* another existing reference, exactly as the SET_ITEM macros do. That actually gives me an idea that wasn't on Serhiy's original list: Py_SET_ATTR(target, value). After all, setting attributes safely from C is the main use case for this, and I think it strengthens the parallel with the SET_ITEM macros on the concrete types. Cheers, Nick.

...

-Barry _______________________________________________ Python-Dev mailing list Python-Dev@python.org https://mail.python.org/mailman/listinfo/python-dev Unsubscribe:

https://mail.python.org/mailman/options/python-dev/ncoghlan%40gmail.com

Barry Warsaw

2:49 p.m.

On Mar 01, 2014, at 08:15 AM, Nick Coghlan wrote:

...

It *is* playing refcounting games - it's decrefing the existing target while stealing a reference to the new target, just like the existing SET_ITEM macros and somewhat like Py_CLEAR (although in that case, it's more obvious that we will never incref NULL).

Okay, but "setting the reference" isn't one of them, which is what I read when I see Py_SETREF. ;)

...

The whole point of this macro is to take an *existing* reference and safely *overwrite* another existing reference, exactly as the SET_ITEM macros do.

That actually gives me an idea that wasn't on Serhiy's original list: Py_SET_ATTR(target, value).

That does seem better. -Barry

emm.odeke＠gmail.com

5:03 p.m.

Sent from my BlackBerry 10 smartphone on the Koodo network. Original Message From: Barry Warsaw Sent: Friday, February 28, 2014 3:50 PM To: python-dev@python.org Subject: Re: [Python-Dev] Poll: Py_REPLACE/Py_ASSIGN/etc On Mar 01, 2014, at 08:15 AM, Nick Coghlan wrote:

...

It *is* playing refcounting games - it's decrefing the existing target while stealing a reference to the new target, just like the existing SET_ITEM macros and somewhat like Py_CLEAR (although in that case, it's more obvious that we will never incref NULL).

Okay, but "setting the reference" isn't one of them, which is what I read when I see Py_SETREF. ;)

...

The whole point of this macro is to take an *existing* reference and safely *overwrite* another existing reference, exactly as the SET_ITEM macros do.

That actually gives me an idea that wasn't on Serhiy's original list: Py_SET_ATTR(target, value).

That does seem better. -Barry

Serhiy Storchaka

18 Mar 18 Mar

11:29 a.m.

26.02.14 11:40, Serhiy Storchaka написав(ла):

...

Let's choose the least confusing names.

See discussions at:

http://bugs.python.org/issue3081 http://bugs.python.org/issue16447 http://bugs.python.org/issue20440 http://comments.gmane.org/gmane.comp.python.devel/145346

Poll results: Py_(X)SETREF: +3 (Antoine, Kristján, Nick) Py_(X)DECREC_REPLACE: +3 (Ryan, Georg, Larry) -2 (Antoine, Kristján) Py_(X)ASSIGN, Py_REF_ASSIGN, Py_(X)REPLACE, Py_(X)STORE, Py_SET_AND_(X)DECREF, Py_(X)DECREF_AND_ASSIGN, Py_ASSIGN_AND_(X)DECREF: -1 (Antoine or Kristján) Py_CLEAR_AND_SET: -2 (Antoine, Kristján)

Barry Warsaw

11:37 a.m.

On Mar 18, 2014, at 08:29 PM, Serhiy Storchaka wrote:

...

Poll results:

Py_(X)SETREF: +3 (Antoine, Kristján, Nick)

+1 -Barry

Georg Brandl

12:05 p.m.

Am 18.03.2014 19:29, schrieb Serhiy Storchaka:

...

26.02.14 11:40, Serhiy Storchaka написав(ла):

...
Let's choose the least confusing names.

See discussions at:

http://bugs.python.org/issue3081 http://bugs.python.org/issue16447 http://bugs.python.org/issue20440 http://comments.gmane.org/gmane.comp.python.devel/145346

Poll results:

Py_(X)SETREF: +3 (Antoine, Kristján, Nick)

Well, now that I know that -1 are counted, -1 to this.

...

Py_(X)DECREC_REPLACE: +3 (Ryan, Georg, Larry) -2 (Antoine, Kristján)

Py_(X)ASSIGN, Py_REF_ASSIGN, Py_(X)REPLACE, Py_(X)STORE, Py_SET_AND_(X)DECREF, Py_(X)DECREF_AND_ASSIGN, Py_ASSIGN_AND_(X)DECREF: -1 (Antoine or Kristján)

Py_CLEAR_AND_SET: -2 (Antoine, Kristján)

Georg

Larry Hastings

12:08 p.m.

On 03/18/2014 12:05 PM, Georg Brandl wrote:

...

Am 18.03.2014 19:29, schrieb Serhiy Storchaka:

...
26.02.14 11:40, Serhiy Storchaka написав(ла):

...
Let's choose the least confusing names.

See discussions at:

http://bugs.python.org/issue3081 http://bugs.python.org/issue16447 http://bugs.python.org/issue20440 http://comments.gmane.org/gmane.comp.python.devel/145346

Poll results:

Py_(X)SETREF: +3 (Antoine, Kristján, Nick) Well, now that I know that -1 are counted, -1 to this.

Me too, -1 to Py_SETREF / Py_XSETREF. //arry/

Nick Coghlan

2:06 p.m.

On 19 Mar 2014 05:10, "Larry Hastings" wrote:

...

On 03/18/2014 12:05 PM, Georg Brandl wrote:

...
Am 18.03.2014 19:29, schrieb Serhiy Storchaka:

...
26.02.14 11:40, Serhiy Storchaka написав(ла):

...
Let's choose the least confusing names.

See discussions at:

http://bugs.python.org/issue3081 http://bugs.python.org/issue16447 http://bugs.python.org/issue20440 http://comments.gmane.org/gmane.comp.python.devel/145346

Poll results:

Py_(X)SETREF: +3 (Antoine, Kristján, Nick)

Well, now that I know that -1 are counted, -1 to this.

Me too, -1 to Py_SETREF / Py_XSETREF.

I think we're going to have to resort to BDFL appeal on this one - consensus seems unlikely at this point. Cheers, Nick.

...

/arry

_______________________________________________ Python-Dev mailing list Python-Dev@python.org https://mail.python.org/mailman/listinfo/python-dev Unsubscribe:

https://mail.python.org/mailman/options/python-dev/ncoghlan%40gmail.com

...

MRAB

2:19 p.m.

On 2014-03-18 21:06, Nick Coghlan wrote:

...

On 19 Mar 2014 05:10, "Larry Hastings" mailto:larry@hastings.org> wrote:

...
On 03/18/2014 12:05 PM, Georg Brandl wrote:

...
Am 18.03.2014 19:29, schrieb Serhiy Storchaka:

...
26.02.14 11:40, Serhiy Storchaka написав(ла):

...
Let's choose the least confusing names.

See discussions at:

http://bugs.python.org/issue3081 http://bugs.python.org/issue16447 http://bugs.python.org/issue20440 http://comments.gmane.org/gmane.comp.python.devel/145346

Poll results:

Py_(X)SETREF: +3 (Antoine, Kristján, Nick)

Well, now that I know that -1 are counted, -1 to this.

Me too, -1 to Py_SETREF / Py_XSETREF.

I think we're going to have to resort to BDFL appeal on this one - consensus seems unlikely at this point.

FWIW, I haven't been following the discussion, but, after a (very) quick look, to me: Py_XDECREF(ptr); ptr = NULL; would be clearer as: Py_CLEAR_REF(ptr); and: Py_XDECREF(ptr); ptr = new_value; would be clearer as: Py_REPLACE_REF(ptr, new_value);

Serhiy Storchaka

2:31 p.m.

26.02.14 11:40, Serhiy Storchaka написав(ла):

...

Let's choose the least confusing names.

See discussions at:

http://bugs.python.org/issue3081 http://bugs.python.org/issue16447 http://bugs.python.org/issue20440 http://comments.gmane.org/gmane.comp.python.devel/145346

Updated poll results. There are two leaders: Py_(X)SETREF (originally proposed by Antoine in issue3081): +4 (Antoine, Kristján, Nick, Barry) -2 (Georg, Larry) Py_(X)DECREC_REPLACE (originally proposed by Victor in issue16447): +3 (Ryan, Georg, Larry) -2 (Antoine, Kristján)

Nick Coghlan

3:26 p.m.

On 19 Mar 2014 07:34, "MRAB" wrote:

...

FWIW, I haven't been following the discussion,

Note that this about correctness, not just clarity - using DECREF on member attributes is not safe, as you may end up exposing a partially destroyed object to other code.

...

but, after a (very) quick look, to me:

Py_XDECREF(ptr); ptr = NULL;

would be clearer as:

Py_CLEAR_REF(ptr);

Already exists as Py_CLEAR (with the correct temp variable usage).

...

and:

Py_XDECREF(ptr); ptr = new_value;

would be clearer as:

Py_REPLACE_REF(ptr, new_value);

That is indeed the one we're aiming to find a suitable name for. Cheers, Nick.

...

_______________________________________________ Python-Dev mailing list Python-Dev@python.org https://mail.python.org/mailman/listinfo/python-dev Unsubscribe:

https://mail.python.org/mailman/options/python-dev/ncoghlan%40gmail.com

3691

Age (days ago)

3711

Last active (days ago)

List overview

Download

20 comments

10 participants

participants (10)

Antoine Pitrou
Barry Warsaw
emm.odeke＠gmail.com
Georg Brandl
Kristján Valur Jónsson
Larry Hastings
MRAB
Nick Coghlan
Ryan Gonzalez
Serhiy Storchaka

Poll: Py_REPLACE/Py_ASSIGN/etc

Kristján Valur Jónsson

Kristján Valur Jónsson

emm.odeke＠gmail.com

MRAB

tags

participants (10)