Strange Python behavior from inappropriate usage of 'is not' comparison?

Question

I (incorrectly?) used 'is not' in a comparison and found this curious behavior:

>>> a = 256
>>> b = int('256')
>>> c = 300
>>> d = int('300')
>>>
>>> a is not b
False
>>> c is not d
True

Obviously I should have used:

>>> a != b
False
>>> c != d
False

But it worked for a long time due to small-valued test-cases until I happened to use a number of 495.

If this is invalid syntax, then why? And shouldn't I at least get a warning?

Answer 1

Don't use is [not] to compare integers; use == and != instead. Even though is works in current CPython for small numbers due to an optimization, it's unreliable and semantically wrong. The syntax itself is valid, but the benefits of a warning (which would have to be checked on every use of is and could be problematic with subclasses of int) are presumably not worth the trouble.

This is covered elsewhere on SO, but I didn't find it just now.

Answer 2

"is" is not a check of equality of value, but a check that two variables point to the same instance of an object.

ints and strings are confusing for this as is and == can happen to give the same result due to how the internals of the language work.

Answer 3

For small numbers, Python is reusing the object instances, but for larger numbers, it creates new instances for them.

See this:

>>> a=256
>>> b=int('256')
>>> c=300       
>>> d=int('300')

>>> id(a)
158013588
>>> id(b)
158013588
>>> id(c)
158151472
>>> id(d)
158151436

which is exactly why a is b, but c isn't d.

Answer 4

For more understanding why this occurs take a look to Python-2.6.5/Objects/intobject.c:78:small_ints array and Python-2.6.5/Objects/intobject.c:1292:_PyInt_Init function in python sources.

Also similar thing occurs with lists:

>>> a = [12]
>>> id_a = id(a)
>>> del(a)
>>> id([1,2,34]) == id_a
True
>>> 

Removed lists are not destroyed. They are reused