How to access outer class from an inner class?

Question

You're trying to access Outer's class instance, from inner class instance. So just use factory-method to build Inner instance and pass Outer instance to it.

class Outer(object):

    def createInner(self):
        return Outer.Inner(self)

    class Inner(object):
        def __init__(self, outer_instance):
            self.outer_instance = outer_instance
            self.outer_instance.somemethod()

        def inner_method(self):
            self.outer_instance.anothermethod()

The methods of a nested class cannot directly access the instance attributes of the outer class.

Note that it is not necessarily the case that an instance of the outer class exists even when you have created an instance of the inner class.

In fact, it is often recommended against using nested classes, since the nesting does not imply any particular relationship between the inner and outer classes.

maybe I'm mad but this seems very easy indeed - the thing is to make your inner class inside a method of the outer class...

def do_sthg( self ):
    ...

def messAround( self ):

    outerClassSelf = self

    class mooble():
        def do_sthg_different( self ):
            ...
            outerClassSelf.do_sthg()

Plus... "self" is only used by convention, so you could do this:

def do_sthg( self ):
    ...

def messAround( outerClassSelf ):

    class mooble():
        def do_sthg_different( self ):
            ...
            outerClassSelf.do_sthg()

It might be objected that you can't then create this inner class from outside the outer class... but this ain't true:

class Bumblebee():

    def do_sthg( self ):
        print "sthg"

    def giveMeAnInnerClass( outerClassSelf ):

        class mooble():
            def do_sthg_different( self ):
                print "something diff\n"
                outerClassSelf.do_sthg()
        return mooble

then, somewhere miles away:

blob = Bumblebee().giveMeAnInnerClass()()
blob.do_sthg_different()    

even push the boat out a bit and extend this inner class (NB to get super() to work you have to change the class signature of mooble to "class mooble( object )"

class InnerBumblebeeWithAddedBounce( Bumblebee().giveMeAnInnerClass() ):
    def bounce( self ):
        print "bounce"

    def do_sthg_different( self ):
        super( InnerBumblebeeWithAddedBounce, self ).do_sthg_different()
        print "and more different"


ibwab = InnerBumblebeeWithAddedBounce()    
ibwab.bounce()
ibwab.do_sthg_different()

later

mrh1997 raised an interesting point about the non-common inheritance of inner classes delivered using this technique. But it seems that the solution is pretty straightforward:

class Fatty():
    def do_sthg( self ):
        pass

    class InnerFatty( object ):
        pass

    def giveMeAnInnerFattyClass(self):
        class ExtendedInnerFatty( Fatty.InnerFatty ):
            pass
        return ExtendedInnerFatty

fatty1 = Fatty()
fatty2 = Fatty()

innerFattyClass1 = fatty1.giveMeAnInnerFattyClass()
innerFattyClass2 = fatty2.giveMeAnInnerFattyClass()

print ( issubclass( innerFattyClass1, Fatty.InnerFatty ))
print ( issubclass( innerFattyClass2, Fatty.InnerFatty ))

I found this.

Tweaked to suite your question:

class Outer(object):
    def some_method(self):
        # do something

    class _Inner(object):
        def __init__(self, outer):
            outer.some_method()
    def Inner(self):
        return _Inner(self)

I’m sure you can somehow write a decorator for this or something

related: What is the purpose of python's inner classes?

You can easily access to outer class using metaclass: after creation of outer class check it's attribute dict for any classes (or apply any logic you need - mine is just trivial example) and set corresponding values:

import six
import inspect


# helper method from `peewee` project to add metaclass
_METACLASS_ = '_metaclass_helper_'
def with_metaclass(meta, base=object):
    return meta(_METACLASS_, (base,), {})


class OuterMeta(type):
    def __new__(mcs, name, parents, dct):
        cls = super(OuterMeta, mcs).__new__(mcs, name, parents, dct)
        for klass in dct.values():
            if inspect.isclass(klass):
                print("Setting outer of '%s' to '%s'" % (klass, cls))
                klass.outer = cls

        return cls


# @six.add_metaclass(OuterMeta) -- this is alternative to `with_metaclass`
class Outer(with_metaclass(OuterMeta)):
    def foo(self):
        return "I'm outer class!"

    class Inner(object):
        outer = None  # <-- by default it's None

        def bar(self):
            return "I'm inner class"


print(Outer.Inner.outer)
>>> <class '__main__.Outer'>
assert isinstance(Outer.Inner.outer(), Outer)

print(Outer().foo())
>>> I'm outer class!
print(Outer.Inner.outer().foo())
>>> I'm outer class!
print(Outer.Inner().outer().foo())
>>> I'm outer class!
print(Outer.Inner().bar())
>>> I'm inner class!

Using this approach, you can easily bind and refer two classes between each other.

I've created some Python code to use an outer class from its inner class, based on a good idea from another answer for this question. I think it's short, simple and easy to understand.

class higher_level__unknown_irrelevant_name__class:
    def __init__(self, ...args...):
        ...other code...
        # Important lines to access sub-classes.
        subclasses = self._subclass_container()
        self.some_subclass = subclasses["some_subclass"]
        del subclasses # Free up variable for other use.

    def sub_function(self, ...args...):
        ...other code...

    def _subclass_container(self):
        _parent_class = self # Create access to parent class.
        class some_subclass:
            def __init__(self):
                self._parent_class = _parent_class # Easy access from self.
                # Optional line, clears variable space, but SHOULD NOT BE USED
                # IF THERE ARE MULTIPLE SUBCLASSES as would stop their parent access.
                #  del _parent_class
        class subclass_2:
            def __init__(self):
                self._parent_class = _parent_class
        # Return reference(s) to the subclass(es).
        return {"some_subclass": some_subclass, "subclass_2": subclass_2}

The main code, "production ready" (without comments, etc.). Remember to replace all of each value in angle brackets (e.g. <x>) with the desired value.

class <higher_level_class>:
    def __init__(self):
        subclasses = self._subclass_container()
        self.<sub_class> = subclasses[<sub_class, type string>]
        del subclasses

    def _subclass_container(self):
        _parent_class = self
        class <sub_class>:
            def __init__(self):
                self._parent_class = _parent_class
        return {<sub_class, type string>: <sub_class>}

Explanation of how this method works (the basic steps):

1. Create a function named _subclass_container to act as a wrapper to access the variable self, a reference to the higher level class (from code running inside the function).

   1. Create a variable named _parent_class which is a reference to the variable self of this function, that the sub-classes of _subclass_container can access (avoids name conflicts with other self variables in subclasses).

   2. Return the sub-class/sub-classes as a dictionary/list so code calling the _subclass_container function can access the sub-classes inside.

2. In the __init__ function inside the higher level class (or wherever else needed), receive the returned sub-classes from the function _subclass_container into the variable subclasses.

3. Assign sub-classes stored in the subclasses variable to attributes of the higher level class.

A few tips to make scenarios easier:

Making the code to assign the sub classes to the higher level class easier to copy and be used in classes derived from the higher level class that have their __init__ function changed:

Insert before line 12 in the main code:

def _subclass_init(self):

Then insert into this function lines 5-6 (of the main code) and replace lines 4-7 with the following code:

self._subclass_init(self)

Making subclass assigning to the higher level class possible when there are many/unknown quantities of subclasses.

Replace line 6 with the following code:

for subclass_name in list(subclasses.keys()):
    setattr(self, subclass_name, subclasses[subclass_name])

Example scenario of where this solution would be useful and where the higher level class name should be impossible to get:

A class, named "a" (class a:) is created. It has subclasses that need to access it (the parent). One subclass is called "x1". In this subclass, the code a.run_func() is run.

Then another class, named "b" is created, derived from class "a" (class b(a):). After that, some code runs b.x1() (calling the sub function "x1" of b, a derived sub-class). This function runs a.run_func(), calling the function "run_func" of class "a", not the function "run_func" of its parent, "b" (as it should), because the function which was defined in class "a" is set to refer to the function of class "a", as that was its parent.

This would cause problems (e.g. if function a.run_func has been deleted) and the only solution without rewriting the code in class a.x1 would be to redefine the sub-class x1 with updated code for all classes derived from class "a" which would obviously be difficult and not worth it.