isinstance is broken

[Carl Banks]

Michele Simionato wrote:
> There are inconsistencies in "isinstance".
> 
> Example 1:
> 
>>>> isinstance(int,object)
> 1
> 
> gives no error message. I don't like it, an error should be
> raised. There is issubclass for that job:

Even if it were wrong--why should it raise an error?  If int were not
an instance of object, then it should just return 0 (or False).


>>>> issubclass(int,object)
> 1

int is both a subclass and an instance of object.

For A to be an instance of B, then B has to be either type(A), or any
base class of type(A).

int's type is type.
object is a base class of type.
Therefore, int is an instance of object.

It's called inheritance.  It's just like C++ and other typical OO
languages: if an object is an instance of a class, it is also an
instance of that class's bases.


> Should I interpret the fact that isinstance(something,object) always returns
> true as a bug or is it intentional ?

Neither.  It's the right behavior.  (I don't think it's manifestly
true, BTW.  I think a C extension could create an object that is not
an instance of object, in which case isinstance(something,object)
would return false.  Someone might want to verify.)


-- 
CARL BANKS