Private Attribute

Introduction

This package provide a way to create the private attribute like "C++" does.

All API

from private_attribute import (PrivateAttrBase, PrivateWrapProxy)      # 1 Import public API

def my_generate_func(obj_id, attr_name):                           # 2 Optional: custom name generator
    return f"_hidden_{obj_id}_{attr_name}"

class MyClass(PrivateAttrBase, private_func=my_generate_func):     # 3 Inherit + optional custom generator
    __private_attrs__ = ['a', 'b', 'c', 'result', 'conflicted_name']  # 4 Must declare all private attrs

    def __init__(self):
        self.a = 1
        self.b = 2
        self.c = 3
        self.result = 42                    # deliberately conflicts with internal names

    # Normal methods can freely access private attributes
    def public_way(self):
        print(self.a, self.b, self.c)

    # Real-world case: method wrapped by multiple decorators
    @PrivateWrapProxy(memoize())                                   # 5 Apply any decorator safely
    @PrivateWrapProxy(login_required())                            # 5 Stack as many as needed
    @PrivateWrapProxy(rate_limit(calls=10))                        # 5
    def expensive_api_call(self, x):                               # First definition (will be wrapped)
        def inner(...):
            return some_implementation(self.a, self.b, self.c, x)
        inner(...)
        return heavy_computation(self.a, self.b, self.c, x)

    @expensive_api_call.non_conflict_attr_name1                    # 6 Easy access to internal names
    @expensive_api_call.non_conflict_attr_name2                    # 6 Easy use when the name has no conflict
    @PrivateWrapProxy(lambda f: f)                                 # 5 dummy wrapper just to restore order
    def expensive_api_call(self, x):                               # Second definition (will be wrapped)
        return heavy_computation(self.a, self.b, self.c, x)

    # Fix decorator order + resolve name conflicts
    @PrivateWrapProxy(expensive_api_call.result.conflicted_name2, expensive_api_call)    # 7 Chain .result to push decorators down
    @PrivateWrapProxy(expensive_api_call.result.conflicted_name1, expensive_api_call)    # 7 Resolve conflict with internal names
    def expensive_api_call(self, x):         # Final real implementation
        return heavy_computation(self.a, self.b, self.c, x)


# ====================== Usage ======================
obj = MyClass()
obj.public_way()                    # prints: 1 2 3

print(hasattr(obj, 'a'))            # False – truly hidden from outside
print(obj.expensive_api_call(10))   # works with all decorators applied

Usage

This is a simple usage about the module:

from private_attribute import PrivateAttrBase

class MyClass(PrivateAttrBase):
    __private_attrs__ = ['a', 'b', 'c']
    def __init__(self):
        self.a = 1
        self.b = 2
        self.c = 3

    def public_way(self):
        print(self.a, self.b, self.c)

obj = MyClass()
obj.public_way()  # (1, 2, 3)

print(hasattr(obj, 'a'))  # False
print(hasattr(obj, 'b'))  # False
print(hasattr(obj, 'c'))  # False

All of the attributes in __private_attrs__ will be hidden from the outside world, and stored by another name.

You can use your function to generate the name. It needs the id of the obj and the name of the attribute:

def my_generate_func(obj_id, attr_name):
    return some_string

class MyClass(PrivateAttrBase, private_func=my_generate_func):
    __private_attrs__ = ['a', 'b', 'c']
    def __init__(self):
        self.a = 1
        self.b = 2
        self.c = 3

    def public_way(self):
        print(self.a, self.b, self.c)

obj = MyClass()
obj.public_way()  # (1, 2, 3)

If the method will be decorated, the property, classmethod and staticmethod will be supported. For the other, you can use the PrivateWrapProxy to wrap the function:

from private_attribute import PrivateAttrBase, PrivateWrapProxy

class MyClass(PrivateAttrBase):
    __private_attrs__ = ['a', 'b', 'c']
    @PrivateWrapProxy(decorator1())
    @PrivateWrapProxy(decorator2())
    def method1(self):
        ...

    @method1.attr_name
    @PrivateWrapProxy(lambda _: _) # use empty function to wrap
    def method1(self):
        ...

    @PrivateWrapProxy(decorator3())
    def method2(self):
        ...

    @method2.attr_name
    @PrivateWrapProxy(lambda _: _)
    def method2(self):
        ...

The PrivateWrapProxy is a decorator, and it will wrap the function with the decorator. When it decorates the method, it returns a _PrivateWrap object. "__getattr__" on the _PrivateWrap object will return the _PrivateWrapParent object, will on _PrivateWrapParent it change itself and return itself.

Both _PrivateWrap and _PrivateWrapParent have the public api result.

Here are the attr name that maybe conflict:

["result", "_result", "_private_result", "_func_list", "__func_list__", "_private_obj", "_private_parent"]

If the attribute name has confilct with the list above, you can use this code:

from private_attribute import PrivateAttrBase, PrivateWrapProxy

class MyClass(PrivateAttrBase):
    __private_attrs__ = ['a', 'b', 'c']
    @PrivateWrapProxy(decorator1())
    @PrivateWrapProxy(decorator2())
    def method1(self):
        ...

    @PrivateWrapProxy(method1.result.conflict_attr_name1, method1) # Use the argument "method1" to save old func
    def method1(self):
        ...

    @PrivateWrapProxy(method1.result.conflict_attr_name2, method1)
    def method1(self):
        ...

    @PrivateWrapProxy(decorator3())
    def method2(self):

Notes

• All of the private attributes class must contain the __private_attrs__ attribute.
• The __private_attrs__ attribute must be a sequence of strings.
• You cannot define the name which in __slots__ to __private_attrs__.
• When you define __slots__ and __private_attrs__ in one class, the attributes in __private_attrs__ can also be defined in the methods, even though they are not in __slots__.
• All of the object that is the instance of the class "PrivateAttrBase" or its subclass are default to be unable to be pickled.
• Finally the attributes' names in __private_attrs__ will be change to a tuple with two hash.
• Finally the _PrivateWrap object will be recoveried to the original object.
• One class defined in another class cannot use another class's private attribute.
• One parent class defined an attribute which not in __private_attrs__ or not a PrivateAttrType instance, the child class shouldn't contain the attribute in its __private_attrs__.

License

MIT