Multiple constructors in python?

Question

Possible Duplicate:
What is a clean, pythonic way to have multiple constructors in Python?

Is it not possible to define multiple constructors in Python, with different signatures? If not, what's the general way of getting around it?

For example, let's say you wanted to define a class City

I'd like to be able to say someCity = City() or someCity = City("Berlin"), where the first just gives a default name value, and the second defines it.

Answer 1

Easiest way is through keyword arguments:

class City():
  def __init__(self, city=None):
    pass

someCity = City(city="Berlin")

This is pretty basic stuff, maybe look at the python docs?

Answer 2

For the example you gave, use default values:

class City:
    def __init__(name="Default City Name"):
        ...
    ...

In general, you have two options:

1) Do if-elif blocks based on the type:

def __init__(name):
    if isinstance(name, str):
        ...
    elif isinstance(name, City):
        ...
    ...

2) Use duck typing --- that is, assume the user of your class is intelligent enough to use it correctly. This is typically the preferred option.

Answer 3

Unlike Java, you cannot define multiple constructors. However, you can define a default value if one is not passed.

def __init__(self, city="Berlin"):
  self.city = city

Answer 4

If your signatures differ only in the number of arguments using default arguments is the right way to do it. If you want to be able to pass in different kinds of argument I would try to avoid the isinstance-based approach mentioned above, instead using keyword arguments. If using just keyword arguments becomes unwieldy you can combine it with classmethods (the bzrlib code likes this approach). This is just a silly example, but I hope you get the idea:

class C(object):

    def __init__(self, fd):
        # Assume fd is a file-like object.
        self.fd = fd

    @classmethod
    def fromfilename(cls, name):
        return cls(open(self.fd, 'rb'))

# Now you can do:
c = C(fd)
# or:
c = C.fromfilename('a filename')

Notice all those classmethods still go through the same init, but using classmethods can be more convenient than having to remember what combinations of keyword arguments to init work.

isinstance is best avoided because python's duck typing makes it hard to figure out what kind of object was actually passed in. For example: if you want to take either a filename or a file-like object you cannot use isinstance(arg, file) because there are many file-like objects that do not subclass file (like the ones returned from urllib, or StringIO, or...). It's usually a better idea to just have the caller tell you explicitly what kind of object was meant, by using different keyword arguments.

Answer 5

Jack M. is right, do it this way:

>>> class City:
...     def __init__(self, city=None):
...         self.city = city
...     def __repr__(self):
...         if self.city:  return self.city
...         return ''
... 
>>> c = City('Berlin')
>>> print c
Berlin
>>> c = City()
>>> print c

>>>