The Descriptor Protocol

Python's dynamic nature extends to modifying objects at runtime. Methods can be added to instances after creation—but doing so requires understanding the descriptor protocol.

The Problem: Adding Methods Dynamically

Consider adding a method to an existing instance:

class MyClass:
    def __init__(self):
        self.value = 10


def power(self, x):
    return self.value ** x


my_instance = MyClass()
my_instance.power = power

Calling my_instance.power(2) raises an error:

TypeError: power() missing 1 required positional argument: 'x'

The function was assigned as an attribute, but Python doesn't automatically pass the instance as the first argument. The function isn't bound to the instance.

How Method Binding Works

When a method is accessed on an instance, Python uses the descriptor protocol to bind it. Any object that defines __get__, __set__, or __delete__ is a descriptor.

Functions are descriptors—they implement __get__. When accessed through a class or instance, __get__ creates a bound method that includes the instance as the first argument.

Manual Binding

The __get__ method can be called directly to bind a function:

# Manually bind the function to the instance
bound_method = power.__get__(my_instance, MyClass)
print(bound_method(2))  # 100 (10 ** 2)

The bound method now knows about my_instance and will pass it as self.

Assigning the Bound Method

The bound method can be assigned to the instance:

my_instance.power = power.__get__(my_instance, MyClass)
print(my_instance.power(3))  # 1000 (10 ** 3)

This demonstrates that methods in Python are objects that can be created, passed around, and assigned like any other value.

Why This Matters

The descriptor protocol underlies several Python features:

• @property: Uses descriptors to create managed attributes
• @classmethod and @staticmethod: Modify how functions are bound
• ORMs: Use descriptors to map object attributes to database columns

Understanding descriptors provides insight into how Python's attribute access actually works.