You can define a function that takes an arbitrary number of keyword (named) arguments by using the double star ** before a parameter name:

def print_kwargs(**kwargs):
    print(kwargs)

When calling the method, Python will construct a dictionary of all keyword arguments and make it available in the function body:

print_kwargs(a="two", b=3)
# prints: "{a: "two", b=3}"

Note that the **kwargs parameter in the function definition must always be the last parameter, and it will only match the arguments that were passed in after the previous ones.

def example(a, **kw):
    print kw

example(a=2, b=3, c=4) # => {'b': 3, 'c': 4}

Inside the function body, kwargs is manipulated in the same way as a dictionary; in order to access individual elements in kwargs you just loop through them as you would with a normal dictionary:

def print_kwargs(**kwargs):
    for key in kwargs:
        print("key = {0}, value = {1}".format(key, kwargs[key])) 

Now, calling print_kwargs(a="two", b=1) shows the following output:

print_kwargs(a = "two", b = 1)
key = a, value = "two"
key = b, value = 1

You can use the star * when writing a function to collect all positional (ie. unnamed) arguments in a tuple:

def print_args(farg, *args):
   print("formal arg: %s" % farg)
   for arg in args:
       print("another positional arg: %s" % arg)

Calling method:

print_args(1, "two", 3)

In that call, farg will be assigned as always, and the two others will be fed into the args tuple, in the order they were received.

You can use a dictionary to assign values to the function's parameters; using parameters name as keys in the dictionary and the value of these arguments bound to each key:

def test_func(arg1, arg2, arg3): # Usual function with three arguments
   print("arg1: %s" % arg1)
   print("arg2: %s" % arg2)
   print("arg3: %s" % arg3)

# Note that dictionaries are unordered, so we can switch arg2 and arg3. Only the names matter.
kwargs = {"arg3": 3, "arg2": "two"}

# Bind the first argument (ie. arg1) to 1, and use the kwargs dictionary to bind the others
test_var_args_call(1, **kwargs) 

Python 3 allows you to define function arguments which can only be assigned by keyword, even without default values. This is done by using star * to consume additional positional parameters without setting the keyword parameters. All arguments after the * are keyword-only (i.e. non-positional) arguments. Note that if keyword-only arguments aren't given a default, they are still required when calling the function.

def print_args(arg1, *args, keyword_required, keyword_only=True):
    print("first positional arg: {}".format(arg1))
    for arg in args:
        print("another positional arg: {}".format(arg))
    print("keyword_required value: {}".format(keyword_required))
    print("keyword_only value: {}".format(keyword_only))
    
print(1, 2, 3, 4) # TypeError: print_args() missing 1 required keyword-only argument: 'keyword_required'
print(1, 2, 3, keyword_required=4) 
# first positional arg: 1
# another positional arg: 2
# another positional arg: 3
# keyword_required value: 4
# keyword_only value: True

The effect of using the * operator on an argument when calling a function is that of unpacking the list or a tuple argument

def print_args(arg1, arg2):
    print(str(arg1) + str(arg2))

a = [1,2]
b = tuple([3,4])

print_args(*a)
# 12
print_args(*b)
# 34

Note that the length of the starred argument need to be equal to the number of the function's arguments.

A common python idiom is to use the unpacking operator * with the zip function to reverse its effects:

a = [1,3,5,7,9]
b = [2,4,6,8,10]

zipped = zip(a,b)
# [(1,2), (3,4), (5,6), (7,8), (9,10)]

zip(*zipped)
# (1,3,5,7,9), (2,4,6,8,10)

To use default values with **kwargs

def fun(**kwargs):
    print kwargs.get('value', 0)

fun()
# print 0
fun(value=1)
# print 1

A common use case for *args in a function definition is to delegate processing to either a wrapped or inherited function. A typical example might be in a class's __init__ method

class A(object):
    def __init__(self, b, c):
        self.y = b
        self.z = c

class B(A):
    def __init__(self, a, *args, **kwargs):
        super(B, self).__init__(*args, **kwargs)
        self.x = a

Here, the a parameter is processed by the child class after all other arguments (positional and keyword) are passed onto - and processed by - the base class.

For instance:

b = B(1, 2, 3)
b.x  # 1
b.y  # 2
b.z  # 3

What happens here is the class B __init__ function sees the arguments 1, 2, 3. It knows it needs to take one positional argument (a), so it grabs the first argument passed in (1), so in the scope of the function a == 1.

Next, it sees that it needs to take an arbitrary number of positional arguments (*args) so it takes the rest of the positional arguments passed in (1, 2) and stuffs them into *args. Now (in the scope of the function) args == [2, 3].

Then, it calls class A's __init__ function with *args. Python sees the * in front of args and "unpacks" the list into arguments. In this example, when class B's __init__ function calls class A's __init__ function, it will be passed the arguments 2, 3 (i.e. A(2, 3)).

Finally, it sets its own x property to the first positional argument a, which equals 1.

Remarks

There a few things to note:

 def func(*args, **kwargs):
     print(args)
     print(kwargs)

 def func(*a, **b):
     print(a)
     print(b)

 def func(*args1, *args2):
 #   File "<stdin>", line 1
 #     def test(*args1, *args2):
 #                      ^
 # SyntaxError: invalid syntax

 def test(**kwargs1, **kwargs2):
 #   File "<stdin>", line 1
 #     def test(**kwargs1, **kwargs2):
 #                       ^
 # SyntaxError: invalid syntax

 def func(a, b, *args, x, y):
     print(a, b, args, x, y)

 func(1, 2, 3, 4, x=5, y=6)
 #>>> 1, 2, (3, 4), 5, 6

 def func(a, b, *, x, y):
     print(a, b, x, y)

 func(1, 2, x=5, y=6)
 #>>> 1, 2, 5, 6

 def test(**kwargs, *args):
 #   File "<stdin>", line 1
 #     def test(**kwargs, *args):
 #                      ^
 # SyntaxError: invalid syntax