Newbies too often do something like (3.2.2, )

>>> id(getattr(x, 'pop')) == id(x.pop)
True

and get confused by the (invalid) result, whereas

>>> a,b=getattr(x, 'pop'),x.pop
>>> id(a)==id(b)
False

works properly. I think we should add a sentence or two or three to the id() doc, such as

Since a newly created argument object is either destroyed or becomes inaccessible before the function returns, *id(obj)* is only useful and valid if *obj* exists prior to the call and therefore after its return.
The value of an expression such as *id(666)== id(667)* is arbitrary and meaningless. The id of the first int object might or might not be reused for the second one.

If we add something at all, it should be much shorter. After all, the current doc already states "unique during its lifetime".

I gave a maximum addition. The first sentence may be enough (or even just the second half of it). What people do not get is that the lifetime of new arg objects is the duration of the function call -- and that the output of id for departed objects is uniquely useless and invalid.

I think this is over-explaining an implementation detail and it makes the docs for id() harder to understand.

Possibly there can be a FAQ entry about identity but the id() function itself is no place to go into the quirks of when new objects are created or destroyed.

Also, "getting confused" is often a step in the learning process -- it can't be taken out entirely.

A FAQ entry sounds good to me.

Suggestion attached.

LGTM

+ created and deleted during execution of the ``id()``

If you want to be more accurate you could say "before and after" instead of "during".

New changeset 8525cc1f342f by Georg Brandl in branch '2.7':
Closes #13203: add a FAQ section about seemingly duplicate id()s.
http://hg.python.org/cpython/rev/8525cc1f342f

New changeset 0d5de993db66 by Georg Brandl in branch '3.3':
Closes #13203: add a FAQ section about seemingly duplicate id()s.
http://hg.python.org/cpython/rev/0d5de993db66