On 1 Mar 2014 01:22, "Barry Warsaw" <barry@python.org> 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 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).

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.


> -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