How to do encapsulation in Python?

Question

What's wrong with this? From objective, and functional standpoints?

import sys

class EncapsulationClass(object):

  def __init__(self):
    self.privates = ["__dict__", "privates", "protected", "a"]
    self.protected = ["b"]

    print self.privates

    self.a = 1
    self.b = 2
    self.c = 3
    pass

  def __getattribute__(self, name):
    if sys._getframe(1).f_code.co_argcount == 0:
      if name in self.privates:
        raise Exception("Access to private attribute \"%s\" is not allowed" % name)
      else:
        return object.__getattribute__(self, name)
    else:
      return object.__getattribute__(self, name)

  def __setattr__(self, name, value):
    if sys._getframe(1).f_code.co_argcount == 0:
      if name in self.privates:
        raise Exception("Setting private attribute \"%s\" is not allowed" % name)
      elif name in self.protected:
        raise Exception("Setting protected attribute \"%s\" is not allowed" % name)
      else:
        return object.__setattr__(self, name, value)
    else:
      return object.__setattr__(self, name, value)


example = EncapsulationClass()

example.a = 10 # Exception: Setting private attribute "a" is not allowed
example.b = 10 # Exception: Setting protected attribute "b" is not allowed
example.c = 10 # example.c == 10

example.__dict__["privates"] # Exception: Setting protected attribute "b" is not allowed

What would actually be wrong with doing something like this?

Is there any better way to achieve encapsulation in Python?

Answer 1

In Mark Lutz's book Learning Python, Fifth edition, he mentioned a way of simulating encapsulation of class level like this:

"""
Created on Sun Oct  4 10:16:30 2020
@author: Mark Lutz
A typical implementation of encapsulation in python,
to use, call：@private(‘var1’, ‘var2’...)
"""

def private(*values):
    def decorator(cls):
        class Proxy:
            def __init__(self, *args, **kwargs):
                self.inst = cls(*args, **kwargs)
            def __call__(self, cls, *args, **kwargs):
                return self.inst
            def __getattr__(self, attr):
                if attr in values:
                    raise AttributeError("Private valueiables are not accessible!")
                else: return getattr(self.inst, attr)
            def __setattr__(self, attr, val):
                # Allow access inside the class
                if attr == 'inst': self.__dict__[attr] = val
                elif attr in values:
                    raise AttributeError("Private valueiables are not accessible!")
                else: setattr(self.inst, attr, val)
            def __str__(self):
                return self.inst.__str__()
        return Proxy
    return decorator

this can be used for class-level encapsulation (e.g.limiting the access of a variable or method in a class).

For module-level encapsulation, however, the only way that I can think of is that you create a file and write the init.py. However if those who writes the client program knows the structure of your file / package, this can still not stop them from importing stuff.

Answer 2

Well, Python does not have encapsulation as a sort of "philosophical" decision, in the same way that we use duck typing a lot. Personally I don't see the point of using private or protected arguments in a Python code.

Speaking of your code, it seems to work fine with the following getters and setters:

def set_a(self, v):
    self.a = v

def get_a(self):
    return self.a

if you make the following modification to your last line of __ getattribute __(self, name):

return object.__getattribute__(self, name)

However, you can use sort of a notion of variable-protecting, if you prefix your private variables with __, as mhawke mentioned. Plus, Daniel's comment points out a limitation of your list arguments. You could keep the protected "get/set" behaviour by adding "private" and "protected"in your private list.

Answer 3

Python has encapsulation - you are using it in your class.

What it doesn't have is access control such as private and protected attributes. However, in Python, there is an attribute naming convention to denote private attributes by prefixing the attribute with one or two underscores, e.g:

self._a
self.__a 

A single underscore indicates to the user of a class that an attribute should be considered private to the class, and should not be accessed directly.

A double underscore indicates the same, however, Python will mangle the attribute name somewhat to attempt to hide it.

class C(object):
    def __init__(self):
        self.a = 123    # OK to access directly
        self._a = 123   # should be considered private
        self.__a = 123  # considered private, name mangled

>>> c = C()
>>> c.a
123
>>> c._a
123
>>> c.__a
Traceback (most recent call last):
  File "<stdin>", line 1, in <module>
AttributeError: 'C' object has no attribute '__a'
>>> c._C__a
123

You can see in the last example that the name was changed from __a to _C__a, although it is still accessible within the class as self.__a.