5.3. OOP Attribute Access Modifiers

• Attributes and methods are always public

• No protected and private keywords

• Private and protected is only by convention [1]

• name - public attribute

• _name - protected attribute (non-public by convention)

• __name - private attribute (name mangling)

• __name__ - system attribute (dunder)

• name_ - avoid name collision with built-ins

In Python, attribute access modifiers are used to control the visibility and accessibility of attributes and methods of a class. There are two main types of attribute access modifiers in Python: public and private.

1. Public attributes and methods: Public attributes and methods are accessible from outside the class. In Python, all attributes and methods are public by default. You can define a public attribute or method by simply declaring it in the class definition.

>>> class MyClass:
...     def __init__(self, x):
...         self.x = x
...
...     def my_method(self):
...         print(self.x)
>>>
>>> obj = MyClass(10)
>>> print(obj.x)
10
>>> obj.my_method()
10

In this example, x and my_method() are public attributes and methods of the MyClass class. They can be accessed from outside the class using the dot notation.

2. Private attributes and methods: Private attributes and methods are not accessible from outside the class. In Python, you can define a private attribute or method by prefixing its name with two underscores (__).

>>> class MyClass:
...     def __init__(self, x):
...         self.__x = x
...
...     def __my_method(self):
...         print(self.__x)
>>>
>>>
>>> obj = MyClass(10)
>>>
>>> print(obj.__x)
Traceback (most recent call last):
AttributeError: 'MyClass' object has no attribute '__x'
>>>
>>> obj.__my_method()
Traceback (most recent call last):
AttributeError: 'MyClass' object has no attribute '__my_method'

In this example, __x and __my_method() are private attributes and methods of the MyClass class. They cannot be accessed from outside the class using the dot notation.

However, it is still possible to access private attributes and methods from outside the class using a special name mangling convention. If you prefix the name of a private attribute or method with a single underscore (_), Python will automatically modify the name to include the class name as a prefix. This makes it possible to access private attributes and methods from outside the class, although it is generally not recommended.

>>> class MyClass:
...     def __init__(self, x):
...         self.__x = x
...
...     def __my_method(self):
...         print(self.__x)
>>>
>>> obj = MyClass(10)
>>> print(obj._MyClass__x)
10
>>> obj._MyClass__my_method()
10

In this example, we access the private attributes and methods of the MyClass class using the modified names _MyClass__x and _MyClass__my_method().

>>> class User:
...     def __init__(self):
...         self.firstname = 'Mark'             # public
...         self.lastname = 'Watney'            # public
...         self._email = 'mwatney@nasa.gov'    # convention: protected
...         self._phone = '+1 (234) 567-8910'   # convention: protected
...         self.__username = 'mwatney'         # convention: private (name mangling)
...         self.__password = 'Ares3'           # convention: private (name mangling)
...         self.type_ = 'admin'                # convention: name collision avoidance (input, type, id, vars, print, sum, max, min)
...         self.id_ = 1                        # convention: name collision avoidance
...         self.__doc__ = 'whatever'           # internals: system field
...         self.__module__ = '__main__'        # internals: system field
...         self.__author__ = 'Mark Watney'     # convention: special fields
...         self.__version__ = '1.0.0'          # convention: special fields

5.3.1. SetUp

>>> from dataclasses import dataclass

5.3.2. Example

>>> class User:
...     def __init__(self):
...         self.firstname = 'Mark'             # public
...         self.lastname = 'Watney'            # public
...         self._email = 'mwatney@nasa.gov'    # convention: protected
...         self._phone = '+1 (234) 567-8910'   # convention: protected
...         self.__username = 'mwatney'         # convention: private (name mangling)
...         self.__password = 'Ares3'           # convention: private (name mangling)
...         self.type_ = 'admin'                # convention: name collision avoidance (input, type, id, vars, print, sum, max, min)
...         self.id_ = 1                        # convention: name collision avoidance
...         self.__doc__ = 'whatever'           # internals: system field
...         self.__module__ = '__main__'        # internals: system field
...         self.__author__ = 'Mark Watney'     # convention: special fields
...         self.__version__ = '1.0.0'          # convention: special fields

>>> mark = User()
>>>
>>> vars(mark)  
{'firstname': 'Mark',
 'lastname': 'Watney',
 '_email': 'mwatney@nasa.gov',
 '_phone': '+1 (234) 567-8910',
 '_User__username': 'mwatney',
 '_User__password': 'Ares3',
 'type_': 'admin',
 'id_': 1,
 '__doc__': 'whatever',
 '__module__': '__main__',
 '__author__': 'Mark Watney',
 '__version__': '1.0.0'}

5.3.3. Public Attribute

• name - public attribute

>>> @dataclass
... class User:
...     firstname: str
...     lastname: str
>>>
>>>
>>> mark = User('Mark', 'Watney')

To print attributes directly:

>>> print(mark.firstname)
Mark
>>>
>>> print(mark.lastname)
Watney

To list all the attributes once again we can use vars():

>>> vars(mark)
{'firstname': 'Mark', 'lastname': 'Watney'}

5.3.4. Protected Attribute

• _attrname - protected attribute (non-public by convention)

>>> @dataclass
... class User:
...     _firstname: str
...     _lastname: str
>>>
>>>
>>> mark = User('Mark', 'Watney')

Python will allow the following statement, however your IDE should warn you "Access to a protected member _firstname of a class":

>>> print(mark._firstname)
Mark
>>>
>>> print(mark._lastname)
Watney

To list all the attributes once again we can use vars():

>>> vars(mark)
{'_firstname': 'Mark', '_lastname': 'Watney'}

5.3.5. Private Attribute

• __attrname - private attribute (name mangling)

>>> @dataclass
... class User:
...     __firstname: str
...     __lastname: str
>>>
>>>
>>> mark = User('Mark', 'Watney')

There are no attributes with names __firstname and __lastname:

>>> print(mark.__firstname)
Traceback (most recent call last):
AttributeError: 'User' object has no attribute '__firstname'
>>>
>>> print(mark.__lastname)
Traceback (most recent call last):
AttributeError: 'User' object has no attribute '__lastname'

To print attributes directly:

>>> print(mark._User__firstname)
Mark
>>>
>>> print(mark._User__lastname)
Watney

To list all the attributes once again we can use vars():

>>> vars(mark)  
{'_User__firstname': 'Mark',
 '_User__lastname': 'Watney'}

5.3.6. Name Mangling

Name mangling is a mechanism which adds the class name to the field name. It is particularly useful when we have an inheritance and the child class is overwriting parent field, which we eventually want to get:

>>> class FormalEnglish:
...     __greeting: str = 'Good Morning'
...     __farewell: str = 'Goodbye'
...     greeting = __greeting
...     farewell = __farewell
>>>
>>>
>>> class SlangEnglish(FormalEnglish):
...     __greeting: str = 'Wassup'
...     __farewell: str = 'Cya'
...     greeting = __greeting
...     farewell = __farewell
>>>
>>>
>>> lang = SlangEnglish()

As expected, when accessing field we will get the latest value. The previous value was overwritten by inheritance.

>>> lang.greeting
'Wassup'
>>>
>>> lang.farewell
'Cya'

However thanks to the name mangling we have an additional access to both FormalEnglish and SlangEnglish fields:

>>> lang._FormalEnglish__greeting
'Good Morning'
>>>
>>> lang._FormalEnglish__farewell
'Goodbye'
>>>
>>> lang._SlangEnglish__greeting
'Wassup'
>>>
>>> lang._SlangEnglish__farewell
'Cya'

Name mangling works for both class variables and instance variables.

5.3.7. Name Collision

• Example colliding names: type_, id_, hash_

>>> type_ = type('myobject')
>>> id_ = id('myobject')
>>> hash_ = hash('myobject')

Example:

>>> class User:
...     def __init__(self, firstname, lastname):
...         self.firstname = firstname
...         self.lastname = lastname
...         self.type_ = type(self)
...         self.id_ = id(self)
...         self.hash_ = hash(self)

5.3.8. System Attributes

• __attrname__ - Current module

• obj.__class__ - Class from which object was instantiated

• obj.__dict__ - Stores instance variables

• obj.__doc__ - Object docstring

• obj.__annotations__ - Object attributes type annotations

• obj.__module__ - Name of a module in which object was defined

>>> @dataclass
... class User:
...     firstname: str
...     lastname: str
>>>
>>>
>>> mark = User('Mark', 'Watney')

>>> mark.__class__
<class '__main__.User'>
>>>
>>> mark.__dict__
{'firstname': 'Mark', 'lastname': 'Watney'}
>>>
>>> mark.__doc__
'User(firstname: str, lastname: str)'
>>>
>>> mark.__annotations__
{'firstname': <class 'str'>, 'lastname': <class 'str'>}
>>>
>>> mark.__module__
'__main__'

5.3.9. Show Attributes

• vars() display obj.__dict__

>>> class User:
...     def __init__(self):
...         self.firstname = 'Mark'
...         self.lastname = 'Watney'
...         self._email = 'mwatney@nasa.gov'
...         self._phone = '+1 (234) 567-8910'
...         self.__username = 'mwatney'
...         self.__password = 'Ares3'
...         self.type_ = 'admin'
...         self.id_ = 1
...         self.__doc__ = 'whatever'
...         self.__module__ = '__main__'
...         self.__author__ = 'Mark Watney'
...         self.__version__ = '1.0.0'
>>>
>>>
>>> mark = User()

All attributes:

>>> vars(mark)  
{'firstname': 'Mark',
 'lastname': 'Watney',
 '_email': 'mwatney@nasa.gov',
 '_phone': '+1 (234) 567-8910',
 '_User__username': 'mwatney',
 '_User__password': 'Ares3',
 'type_': 'admin',
 'id_': 1,
 '__doc__': 'whatever',
 '__module__': '__main__',
 '__author__': 'Mark Watney',
 '__version__': '1.0.0'}

Public attributes:

>>> def get_public_attributes(obj):
...     return {attrname: attrvalue
...             for attrname in dir(obj)
...             if (attrvalue := getattr(mark, attrname))
...             and not callable(attrvalue)
...             and not attrname.startswith('_')}
>>>
>>>
>>> get_public_attributes(mark)
{'firstname': 'Mark', 'id_': 1, 'lastname': 'Watney', 'type_': 'admin'}

Protected attributes:

>>> def get_protected_attributes(obj):
...     return {attrname: attrvalue
...             for attrname in dir(obj)
...             if (attrvalue := getattr(obj, attrname))
...             and not callable(attrvalue)
...             and attrname.startswith('_')
...             and not attrname.startswith(f'_{obj.__class__.__name__}_')
...             and not attrname.endswith('_')}
>>>
>>>
>>> get_protected_attributes(mark)
{'_email': 'mwatney@nasa.gov', '_phone': '+1 (234) 567-8910'}

Private attributes:

>>> def get_private_attributes(obj):
...     return {attrname: attrvalue
...             for attrname in dir(obj)
...             if (attrvalue := getattr(obj, attrname))
...             and not callable(attrvalue)
...             and attrname.startswith(f'_{obj.__class__.__name__}_')}
>>>
>>>
>>> get_private_attributes(mark)
{'_User__password': 'Ares3', '_User__username': 'mwatney'}

System attributes:

>>> def get_system_attributes(obj):
...     return {attrname: attrvalue
...             for attrname in dir(obj)
...             if (attrvalue := getattr(obj, attrname))
...             and not callable(attrvalue)
...             and attrname.startswith('__')
...             and attrname.endswith('__')}
>>>
>>>
>>> get_system_attributes(mark)  
{'__author__': 'Mark Watney',
 '__dict__': {...},
 '__doc__': 'whatever',
 '__module__': '__main__',
 '__version__': '1.0.0'}

5.3.10. References

[1]

https://docs.python.org/3/tutorial/classes.html#private-variables

5.3.11. Assignments

Code 5.21. Solution

"""
* Assignment: OOP AttributeAccessModifiers Dataclass
* Complexity: easy
* Lines of code: 5 lines
* Time: 3 min

English:
    1. Modify dataclass `User` to add attributes:
        a. Public: `firstname`, `lastname`
        b. Protected: `email`, `phone`
        c. Private: `username`, `password`
    2. Run doctests - all must succeed

Polish:
    1. Zmodyfikuj dataclass `User` aby dodać atrybuty:
        a. Publiczne: `firstname`, `lastname`
        b. Chronione: `email`, `phone`
        c. Prywatne: `username`, `password`
    2. Uruchom doctesty - wszystkie muszą się powieść

Tests:
    >>> import sys; sys.tracebacklimit = 0
    >>> from inspect import isclass

    >>> assert isclass(User)
    >>> assert hasattr(User, '__annotations__')

    >>> assert 'firstname' in User.__dataclass_fields__
    >>> assert 'lastname' in User.__dataclass_fields__
    >>> assert '_phone' in User.__dataclass_fields__
    >>> assert '_email' in User.__dataclass_fields__
    >>> assert '_User__username' in User.__dataclass_fields__
    >>> assert '_User__password' in User.__dataclass_fields__
"""
from dataclasses import dataclass


# Public attributes: `firstname`, `lastname`
# Protected attributes: `email`, `phone`
# Private attributes: `username`, `password`
# type: type[User]
@dataclass
class User:
    pass

Code 5.22. Solution

"""
* Assignment: OOP AttributeAccessModifiers Init
* Complexity: easy
* Lines of code: 6 lines
* Time: 5 min

English:
    1. Modify dataclass `User` to add attributes:
        a. Public: `firstname`, `lastname`
        b. Protected: `email`, `phone`
        c. Private: `username`, `password`
    2. Do not use `dataclass`
    3. Run doctests - all must succeed

Polish:
    1. Zmodyfikuj dataclass `User` aby dodać atrybuty:
        a. Publiczne: `firstname`, `lastname`
        b. Chronione: `email`, `phone`
        c. Prywatne: `username`, `password`
    2. Nie używaj `dataclass`
    3. Uruchom doctesty - wszystkie muszą się powieść

Tests:
    >>> import sys; sys.tracebacklimit = 0
    >>> from inspect import isclass

    >>> assert isclass(User)

    >>> result = User(
    ...     firstname='Mark',
    ...     lastname='Watney',
    ...     email='mwatney@nasa.gov',
    ...     phone='+1 (234) 555 0000',
    ...     username='mwatney',
    ...     password='Ares3',
    ... )

    >>> assert hasattr(result, 'firstname')
    >>> assert hasattr(result, 'lastname')
    >>> assert hasattr(result, '_email')
    >>> assert hasattr(result, '_phone')
    >>> assert hasattr(result, '_User__username')
    >>> assert hasattr(result, '_User__password')
"""


# Public attributes: `firstname`, `lastname`
# Protected attributes: `email`, `phone`
# Private attributes: `username`, `password`
# type: type[User]
class User:
    def __init__(self):
        ...

Code 5.23. Solution

"""
* Assignment: OOP AttributeAccessModifiers Members
* Complexity: easy
* Lines of code: 3 lines
* Time: 8 min

English:
    1. Extract from class `User` attribute names and their values:
        a. Define `protected: dict` with protected attributes
        b. Define `private: dict` with private attributes
        c. Define `public: dict` with public attributes
    2. Run doctests - all must succeed

Polish:
    1. Wydobądź z klasy `User` nazwy atrybutów i ich wartości:
        a. Zdefiniuj `protected: dict` z atrybutami chronionymi (protected)
        b. Zdefiniuj `private: dict` z atrybutami prywatnymi (private)
        c. Zdefiniuj `public: dict` z atrybutami publicznymi (public)
    2. Uruchom doctesty - wszystkie muszą się powieść

Tests:
    >>> import sys; sys.tracebacklimit = 0

    >>> assert type(public) is dict
    >>> assert all(type(k) is str for k,v in public.items())
    >>> assert all(type(v) is str for k,v in public.items())

    >>> assert type(protected) is dict
    >>> assert all(type(k) is str for k,v in protected.items())
    >>> assert all(type(v) is str for k,v in protected.items())

    >>> assert type(private) is dict
    >>> assert all(type(k) is str for k,v in private.items())
    >>> assert all(type(v) is str for k,v in private.items())

    >>> assert len(public) > 0, \
    'public: list[dict] must not be empty'

    >>> assert len(protected) > 0, \
    'protected: list[dict] must not be empty'

    >>> assert len(private) > 0, \
    'private: list[dict] must not be empty'

    >>> public
    {'firstname': 'Mark', 'lastname': 'Watney'}

    >>> protected
    {'_email': 'mwatney@nasa.gov', '_phone': '+1 (234) 555 1337'}

    >>> private
    {'_User__username': 'mwatney', '_User__password': 'Ares3'}

"""
from dataclasses import dataclass


@dataclass
class User:
    def __init__(self, firstname, lastname, email, phone, username, password):
        self.firstname = firstname
        self.lastname = lastname
        self._email = email
        self._phone = phone
        self.__username = username
        self.__password = password


DATA = User(
    firstname='Mark',
    lastname='Watney',
    email='mwatney@nasa.gov',
    phone='+1 (234) 555 1337',
    username='mwatney',
    password='Ares3',
)

# All public attributes and their values
# type: dict[str,float|str]
public = ...

# All protected attributes and their values
# type: dict[str,float|str]
protected = ...

# All private attributes and their values
# type: dict[str,str]
private = ...