curiosity about the nature of identity (in python)

[Bruno Desthuilliers]

James Stroud a écrit :
> Hello all,
> 
> What /is/ identity in python?

A unique identifier associated with each and every object in the 
process. What exactly is this identifier is left to the implementation - 
FWIW and IIRC, CPython uses the memory address of the C 'object' 
datastructure.

> For example, we can always count on
> 
> py> None is None
> True

Yes. None is a singleton.

> But I have noticed that this works for strings:
> 
> py> "none" is "none"
> True
 >
> and, for example, integers:
> 
> py> 42 is 42
> True

This is an implementation detail of CPython - which does some caching, 
and is by no way specified by the language. You should *never* rely on 
this. FWIW, try:

a = 3900000
b = 3900000
a is b

> And I have noticed that this works for equivalent expressions:
> 
> py> 42 is (40 + 2)
> True
> 
> So, I'm guessing that the integer 42, or the string "none" is cached 
> somewhere

Bingo !-)

> and python looks to see if it is in the cache when evaluating 
> 'is'. 

I don't think the evaluation of 'is' as anything to do with this.

> My guess is supported with this test:
> 
> py> id(42)
> 149679044
> py> id(40+2)
> 149679044
> py> id(7*6)
> 149679044

Now what about this:

>>> id(600)
134745616
>>> id(601)
134745616

id of an object is unique *for the lifetime of this object*. Nothing 
prevents it from being reused later.

> So, I guess my question is to what extent is this equivalency 
> implementation dependent?

cf above

> Is this equivalency a requirement for a 
> legitimate python implementation?

cf above

> Won't these checks slow down 
> evaluation of 'is' considerably?

I really doubt that comparing memory addresses could be slow...

> Does '==' ever fall back and use 'is' 
> (or 'id') for evaluation?

Seems so :

 >>> class Foo(object): pass
...
 >>> f1 = Foo()
 >>> f2 = Foo()
 >>> f3 = f1
 >>> f1 == f2
False
 >>> f1 == f3
True