Question

I have the following (obviously simplified) class

class A(object)
    def __init__(self, a):
        self.a = a
        self.b = 'b'
        # Other class attributes are added


class B(list):
    """
    Some customization of this class...
    """
    pass

BB = B([A(i) for i in range(10)])

I want to do

B.a

and get a list of all the a attributes from every contained item in B. I know in order to do this, I need to overwrite __getattr__, but I’m not sure the best way to implement this. This needs to be generic as B doesn’t know any of the attributes of A that may need to be accessed.

Can someone offer some advice on the implementation of this idea?

Answer 1

General Solution:

If you wish for this to work across the board, then you can override __getattr__() as you thought:

class A(object):
    def __init__(self, a):
        self.a = a
        self.b = a-1

class B(list):
    """
    Some customization of this class...
    """
    def __getattr__(self, name):
        return (getattr(item, name) for item in self)

bb = B([A(i) for i in range(10)])
print(list(bb.a))
print(list(bb.b))

Giving us:

[0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
[-1, 0, 1, 2, 3, 4, 5, 6, 7, 8]

Note that __getattr__() only gets called if the attribute doesn’t already exist. So, if you set bb.b to another value, you will get that instead:

bb = B([A(i) for i in range(10)])
bb.b = 5
print(list(bb.a))
print(bb.b)

Gives us:

[0, 1, 2, 3, 4, 5, 6, 7, 8, 9]
5

Example to show the lack of need for B to know about it’s contents:

>>> import datetime
>>> b = B([datetime.date(2012, 1, 1), datetime.date(2012, 2, 2), datetime.date(2012, 3, 3)])
>>> list(b.month)
[1, 2, 3]

Original Answer:

The easiest way to do this is with a generator expression.

class B(list):
    """
    Some customization of this class...
    """
@property
def a(self):
    return (item.a for item in self)

This generator expression is the equivalent of:

@property
def a(self):
    for item in self:
        yield item.a

I also used the property() builtin as a decorator to make B.a act as an attribute rather than a function.

We can then do:

bb = B([A(i) for i in range(10)])
print(list(bb.a))

and get:

[0, 1, 2, 3, 4, 5, 6, 7, 8, 9]

You could use a list comprehension ([item.a for item in self]) if you definitely wanted a list rather than an iterator, but generally the iterator is more useful, and can be easily made into a list (as shown above).

Note you could also do this even more simply by assigning the generator expression:

class B(list):
        """
        Some customization of this class...
        """
    def __init__(self, *args):
        super(B, self).__init__(*args)
        self.a = (item.a for item in self)

However, this means the generator will be exhausted after the first use, so I would advise against it.