How can I call a subclass method from parent class ?

[bonono]

Hi,

Suppose my class definition is like this :

class A:
name = "A"

@classmethod
def foo(cls):
cls.__super.foo()
cls.bar()

@classmethod
def bar(cls):
print cls.name

class B(A):
name = "B"

class C(B):
name = "C"

What I want is

C.foo() prints 'ABC'
B.foo() prints 'BC'
A.foo() prints 'A'

But when I call C.foo(), it said

AttributeError: class C has no attribute '_A__super'

How should this be coded ?

[Jason Lai]

If you use a newstyle class, e.g. class A(object), then you can get the
superclass with cls.__base__. You could also use super(cls,cls),
although note that it returns a <super> object that isn't exactly the
same thing as a class -- but good enough for just accessing attributes.

Make sure to check that your superclass isn't <object>, otherwise it'll
complain about <object> not having a foo attribute. __base__ is
probably easier for this purpose. Also be careful with multiple
inheritance.

No such thing as __super though.

[bonono]

thanks, it works. Though I don't quite understand what super(cls,cls)
returns, and it doesn't work if I do a super(cls,cls).foo(). But
cls.__base__.foo() do the trick.

thankfully, I don't have multiple inheritance.

Jason Lai wrote:

If you use a newstyle class, e.g. class A(object), then you can get the
superclass with cls.__base__. You could also use super(cls,cls),
although note that it returns a <super> object that isn't exactly the
same thing as a class -- but good enough for just accessing attributes.

Make sure to check that your superclass isn't <object>, otherwise it'll
complain about <object> not having a foo attribute. __base__ is
probably easier for this purpose. Also be careful with multiple
inheritance.

No such thing as __super though.

[Steve Holden]

bo****@gmail.com wrote:

Jason Lai wrote:

If you use a newstyle class, e.g. class A(object), then you can get the
superclass with cls.__base__. You could also use super(cls,cls),
although note that it returns a <super> object that isn't exactly the
same thing as a class -- but good enough for just accessing attributes.

Make sure to check that your superclass isn't <object>, otherwise it'll
complain about <object> not having a foo attribute. __base__ is
probably easier for this purpose. Also be careful with multiple
inheritance.

No such thing as __super though.

thanks, it works. Though I don't quite understand what super(cls,cls)
returns, and it doesn't work if I do a super(cls,cls).foo(). But
cls.__base__.foo() do the trick.

thankfully, I don't have multiple inheritance.

In point of fact super() is most useful when you *do* have multiple
inheritance. The classic example is when classes B and C have a common
supertype A, and then class D is a subclass of both B and C.

Under these circumstances (the so-called "diamond-shaped inheritance
graph") it can be problematic to ensure that each superclass's methods
are called exactly once when methods are being extended rather than
overridden: should a D call B's method, which then calls A's, and if so
how does a D ensure that C's method gets called without it also calling
A's method.

Calls to super() are used to effectively place a linearised oredering on
the superclasses to ansure that the diamond-shaped inheritance pattern
is correctly handled to give the correct method resolution order.

regards
Steve
--
Steve Holden [phone] [phone]
Holden Web LLC www.holdenweb.com
PyCon TX 2006 www.python.org/pycon/

[bonono]

Thanks for the explanation but some how my code fail and since I don't
need multiple inheritance for the moment, I would settle for the not so
clean version.

The documentation of super is not very clear to me too. As seen in my
code, I am using classmethod which may cause some problem.

Steve Holden wrote:

In point of fact super() is most useful when you *do* have multiple
inheritance. The classic example is when classes B and C have a common
supertype A, and then class D is a subclass of both B and C.

Under these circumstances (the so-called "diamond-shaped inheritance
graph") it can be problematic to ensure that each superclass's methods
are called exactly once when methods are being extended rather than
overridden: should a D call B's method, which then calls A's, and if so
how does a D ensure that C's method gets called without it also calling
A's method.

Calls to super() are used to effectively place a linearised oredering on
the superclasses to ansure that the diamond-shaped inheritance pattern
is correctly handled to give the correct method resolution order.

regards
Steve
--
Steve Holden [phone] [phone]
Holden Web LLC www.holdenweb.com
PyCon TX 2006 www.python.org/pycon/