86
votes

Suppose code like this:

class Base:
    def start(self):
        pass
    def stop(self)
        pass

class A(Base):
    def start(self):
        ... do something for A
    def stop(self)
        .... do something for A

class B(Base):
    def start(self):

    def stop(self):

a1 = A(); a2 = A()
b1 = B(); b2 = B()

all = [a1, b1, b2, a2,.....]

Now I want to call methods start and stop (maybe also others) for each object in the list all. Is there any elegant way for doing this except of writing a bunch of functions like

def start_all(all):
    for item in all:
        item.start()

def stop_all(all):
7
a bunch of function is two function? - SilentGhost
This is unrelated to your question, but there's no obvious reason in your example why you need that base class. Python will be very happy to let you have a list of unrelated objects and so long as they all have 'start' and 'stop' methods you can still iterate through them calling the methods. - Duncan
Your defining Base with useless methods and then defining behaviour in A and B reflects a poor design in Python. Rather than using an abstract base class, you can just define A and B and use them interchangeably insofar as they share an interface. Your current way of doing things creates a useless class, which is just extra stuff you don't need. - Mike Graham

7 Answers

6
votes

The *_all() functions are so simple that for a few methods I'd just write the functions. If you have lots of identical functions, you can write a generic function:

def apply_on_all(seq, method, *args, **kwargs):
    for obj in seq:
         getattr(obj, method)(*args, **kwargs)

Or create a function factory:

def create_all_applier(method, doc=None):
    def on_all(seq, *args, **kwargs):
        for obj in seq:
            getattr(obj, method)(*args, **kwargs)
    on_all.__doc__ = doc
    return on_all

start_all = create_all_applier('start', "Start all instances")
stop_all = create_all_applier('stop', "Stop all instances")
...
189
votes

This will work

all = [a1, b1, b2, a2,.....]

map(lambda x: x.start(),all)    

simple example

all = ["MILK","BREAD","EGGS"]
map(lambda x:x.lower(),all)
>>>['milk','bread','eggs']

and in python3

all = ["MILK","BREAD","EGGS"]
list(map(lambda x:x.lower(),all))
>>>['milk','bread','eggs']
34
votes

It seems like there would be a more Pythonic way of doing this, but I haven't found it yet.

I use "map" sometimes if I'm calling the same function (not a method) on a bunch of objects:

map(do_something, a_list_of_objects)

This replaces a bunch of code that looks like this:

 do_something(a)
 do_something(b)
 do_something(c)
 ...

But can also be achieved with a pedestrian "for" loop:

  for obj in a_list_of_objects:
       do_something(obj)

The downside is that a) you're creating a list as a return value from "map" that's just being throw out and b) it might be more confusing that just the simple loop variant.

You could also use a list comprehension, but that's a bit abusive as well (once again, creating a throw-away list):

  [ do_something(x) for x in a_list_of_objects ]

For methods, I suppose either of these would work (with the same reservations):

map(lambda x: x.method_call(), a_list_of_objects)

or

[ x.method_call() for x in a_list_of_objects ]

So, in reality, I think the pedestrian (yet effective) "for" loop is probably your best bet.

23
votes

The approach

for item in all:
    item.start()

is simple, easy, readable, and concise. This is the main approach Python provides for this operation. You can certainly encapsulate it in a function if that helps something. Defining a special function for this for general use is likely to be less clear than just writing out the for loop.

7
votes

maybe map, but since you don't want to make a list, you can write your own...

def call_for_all(f, seq):
    for i in seq:
        f(i)

then you can do:

call_for_all(lamda x: x.start(), all)
call_for_all(lamda x: x.stop(), all)

by the way, all is a built in function, don't overwrite it ;-)

6
votes

Starting in Python 2.6 there is a operator.methodcaller function.

So you can get something more elegant (and fast):

from operator import methodcaller

map(methodcaller('method_name'), list_of_objects)
3
votes

Taking @Ants Aasmas answer one step further, you can create a wrapper that takes any method call and forwards it to all elements of a given list:

class AllOf:
    def __init__(self, elements):
        self.elements = elements
    def __getattr__(self, attr):
        def on_all(*args, **kwargs):
            for obj in self.elements:
                getattr(obj, attr)(*args, **kwargs)
        return on_all

That class can then be used like this:

class Foo:
    def __init__(self, val="quux!"):
        self.val = val
    def foo(self):
        print "foo: " + self.val

a = [ Foo("foo"), Foo("bar"), Foo()]
AllOf(a).foo()

Which produces the following output:

foo: foo
foo: bar
foo: quux!

With some work and ingenuity it could probably be enhanced to handle attributes as well (returning a list of attribute values).