问题
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?
回答1:
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.
回答2:
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.
回答3:
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.
回答4:
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!
来源:https://stackoverflow.com/questions/10973764/python-initialization