Python initialization

Question

I have this code:

def __init__(self, a, b, c, d...):
   self.a = a
   self.b = b
   etc

I'm thinking of replacing it with:

def __init__(self, a, b, c, d...):
   args=locals()
   for key in args:
     self.__dict__[key] = args[key]

Is this a bad idea? Are there any better ways to do this?

Answer 1

Another alternative for Python3 users: the init decorator from drastic.

@init
def __init__(self, a, b, c, d, e, f, g, h):
    # nothing to write here!

Answer 2

Always consider readability over clever design. Is the replacement code easier to read? I would probably just leave it. Remember that simple is better than complex. As the ThiefMaster said, the explicit assignments are more readable.

Answer 3

From the Zen of Python: Simple is better than complex. and Readability counts.

The explicit assignments are much more readable and simpler than using magic to set the values.

Answer 4

Building on @ThiefMaster's comment about **kwargs:

If you are taking in 20 arguments, it might make more sense to require your users to send arguments via keyword instead of position: with 20 arguments, there is a decent chance that someone using your code will get the arguments in the wrong order.

Consider only accepting kwargs while having a predefined list of keys you want to accept and raising a ValueError if you don't receive them. So you could use **kwargs and then check that everything is there. E.g.

INITIAL_ARGS = set(['a','b','c','d','e'...])

def __init__(self, **kwargs):
    if not INITIAL_ARGS.issubset(set(kwargs.iterkeys())):
        raise ValueError("Class <myclass> requires 20 keyword arguments"
                          "only given %d" % len(kwargs))
    self.__dict__.update(kwargs)

Not sure whether this is more or less Pythonic than your original, but it seems like it would save a ton of time later on when trying to figure out why someone using your code might be getting strange errors.