How can I automatically assign __init__ arguments to attributes?

Question

Is there a way to automatically bind to self (some of) the arguments of the __init__ method?

For example, suppose we have:

class Person:
    def __init__(self, name, age, address):
        self.name = name
        self.age = age
        self.address = address
        ...

could a lazy_init decorator be created that allows doing it this way instead?

class Person:
    @lazy_init
    def __init__(self, name, age, address):
        ...

Does any similar technique exist? Or is there a good reason not to try?

Imprest

Dataclasses in Python 3.7 will do exactly this. python.org/dev/peps/pep-0557 – Acred

Answer 1

For future reference for all others who encounter this question in the future (Python 3.7 or above): There is the dataclass decorator that does exactly this.

Example

from dataclasses import dataclass

@dataclass
class Person:
    name
    age
    address

# other methods...

Katabolism

Answer 2

It's definitely possible, here's a somewhat naive implementation:

from functools import wraps

def lazy_init(init):
    import inspect
    arg_names = inspect.getargspec(init)[0]

    @wraps(init)
    def new_init(self, *args):
        for name, value in zip(arg_names[1:], args):
            setattr(self, name, value)
        init(self, *args)

    return new_init

class Person:
    @lazy_init
    def __init__(self, name, age):
        pass

p = Person("Derp", 13)
print p.name, p.age

Once you start having something besides attributes that map to properties, you're going to run into trouble though. You'll need at least some way to specify which args to initialize as properties... at which point it'll just become more hassle than it's worth.

Bina

Thanks. I realize that if the __init__ method is sophisticated it wouldn't be worth it, but the simple ones are so widespread that I think it would be nice to have something like that at hand. – Sheeting

requires changes for keyword arguments – Decompose

Answer 3

You can either use data classes as of Python 3.7 like this:

from dataclasses import dataclass

@dataclass
class MyClass:
    var_1: str
    var_2: float

Or you can make use of the store_attr() method from the fastcore lib like this:

from fastcore.utils import store_attr

class MyClass:
    def __init__(self, var_1, var_2):
        store_attr()

Both result in the equivalent init method:

def __init__(self, var_1, var_2):
    self.var_1 = var_1
    self.var_2 = var_2

Felske

The dataclass approach seems sufficient enough for most cases :-) – Halophyte

Answer 4

Here it goes, in simplest form:

def lazy_init(*param_names):
  def ret(old_init):
    def __init__(self, *args, **kwargs):
      if len(args) > len(param_names):
        raise TypeError("Too many arguments")
      for k in kwargs:
        if k not in param_names:
          raise TypeError("Arg %r unexpected" % k)
      for par, arg in zip(param_names, args):
        setattr(self, par, arg)
      for par, arg in kwargs.items():
        setattr(self, par, arg)
      old_init(*args, **kwargs)
    return __init__
    #
  return ret


class Q(object):
  @lazy_init("a", "b")
  def __init__(self, *args, **kwargs):
    print "Original init"

>>> q = Q(1, 2)
Original init
>>> q.a, q.b
(1, 2)

Consider making a class decorator to cover __str__ too.

Dumpish

Answer 5

initify.py created by me :) it does exactly what you'd expect.

class Animal:
   @init_args(exclude=["name"])
   def __init__(self, name):
       pass

or with default values:

class Animal:
   @init_args
   def __init__(self, name="Default name"):
       pass

or even excluding inheritance:

class Dog(Animal):
   @init_args(exclude=["name"])
   def __init__(self, species):
       pass

https://github.com/prankymat/initify.py

Dinette