super() and type()

[Chris Mellon]

I see super documented, and in use, as below (from the Python documentation)

class C(B):
def meth(self, arg):
super(C, self).meth(arg)

I'd like to not write C all the time, so is there any problem with writing:

class C(B):
def meth(self, arg):
super(type(self), self).meth(arg)
This seems to work in practice but I don't see it used anywhere and
I'm worried what I might be missing.

This was especially brought to my attention because pylint flags the
second usage as invalid, and I'm not sure if it should be considered a
false positive or not.

[Carl Banks]

Chris Mellon wrote:

I see super documented, and in use, as below (from the Python documentation)

class C(B):
def meth(self, arg):
super(C, self).meth(arg)

I'd like to not write C all the time, so is there any problem with writing:

class C(B):
def meth(self, arg):
super(type(self), self).meth(arg)
This seems to work in practice but I don't see it used anywhere and
I'm worried what I might be missing.

This was especially brought to my attention because pylint flags the
second usage as invalid, and I'm not sure if it should be considered a
false positive or not.

PyLint is right. Try running this:
class A(object):
def meth(self,arg):
print "hello"

class B(A):
def meth(self,arg):
super(type(self), self).meth(arg)

class C(B):
def meth(self,arg):
super(type(self), self).meth(arg)

print C().meth(1)
The reason this type(self) returns the type of the object (which
doesn't change), NOT the type of the class it was defined in. That is,
type(self) returns C even in the function B.meth. Since C's superclass
is B, B.meth ends up calling B.meth again, and you get infinite
recursion.

Unfortunately, short of hackery, you're stuck with having to write out
super(C,self).
Carl Banks

[Rob Williscroft]

Chris Mellon wrote in
news:ma**************************************@pyth on.org in
comp.lang.python:

I see super documented, and in use, as below (from the Python
documentation)

class C(B):
def meth(self, arg):
super(C, self).meth(arg)

I'd like to not write C all the time, so is there any problem with
writing:

class C(B):
def meth(self, arg):
super(type(self), self).meth(arg)
This seems to work in practice but I don't see it used anywhere and
I'm worried what I might be missing.

Have you considered what happens if somebody writes:

class D(C):
def meth(self, arg):
super( D, self ).meth( arg )

>
This was especially brought to my attention because pylint flags the
second usage as invalid, and I'm not sure if it should be considered a
false positive or not.

Nope, the type argument to super tells it where you are in the
type hierarchy, type(self) is always the top of the hierarchy.

Rob.
--
http://www.victim-prime.dsl.pipex.com/