Introduction to Dunder Methods

Special methods in Python—widely referred to as dunder methods (short for "double underscore" methods) or magic methods—are the foundational building blocks of the Python Data Model. They allow user-defined classes to hook directly into the Python language runtime, enabling instances to exhibit native behaviors such as iteration, context management, indexing, slicing, callable invocation, arithmetic operations, and string representation.

When you execute expressions such as len(my_obj), x + y, with my_obj:, or print(my_obj), Python does not invoke ad-hoc runtime checks; instead, it delegates these high-level language constructs to well-defined internal protocol methods: my_obj.__len__(), type(x).__add__(x, y), my_obj.__enter__(), and type(my_obj).__str__(my_obj).

Understanding how dunder methods work at both the language and interpreter levels allows you to write idiomatic, elegant, and highly expressive Python frameworks.

1. The Python Data Model & Special Method Lookup

The Python Data Model defines a formal interface that objects can implement to interact with built-in protocols. In Python's C API (CPython), classes have predefined C-level struct slots (such as tp_new, tp_init, tp_repr, tp_call, tp_as_number, and tp_as_sequence). When you define a dunder method in Python, CPython populates these corresponding slots for blazing-fast runtime dispatch.

The Class Lookup Rule (Critical Nuance)

A foundational architectural design decision in Python is that special methods are almost always looked up on the class (the type), not on the instance itself.

If you dynamically assign a dunder method to an instance dictionary (self.__len__ = lambda: 5), built-in operations such as len(self) will not trigger it. The interpreter bypasses instance.__dict__ for special methods to guarantee performance and avoid infinite recursion within the metaclass layer.

2. Object Lifecycle: __new__ vs __init__ vs __del__

Every Python object undergoes three distinct lifecycle phases: Allocation/Creation, Initialization, and Destruction.

Customizing Instance Allocation with __new__

Because __new__ runs before self exists, it is a static method (though implicitly marked, taking cls as its first argument). It is commonly used for:

1. 1
   Subclassing immutable types like int, str, or tuple.
2. 2
   Implementing the Singleton pattern.
3. 3
   Metaprogramming and factory dispatch.

3. Creating Immutable Data Types Subclassing Built-ins

When subclassing immutable objects like str or tuple, modifying self inside __init__ is impossible because the value has already been frozen during memory allocation. Modifying values requires intercepting __new__:

4. Object Destruction with __del__ (Finalizers)

__del__ is called when an object's reference count drops to zero or during garbage collection cycle reclamation.

5. Overview of Core Special Method Categories

Python organizes dunder methods into distinct functional protocols:

6. Summary and Best Practices

1. 1
   Protocol Over Class Inheritance: Python adheres to duck typing. If an object implements __iter__ and __next__, it is an iterator; if it implements __getitem__ and __len__, it is a sequence.
2. 2
   Never Invent Custom Dunder Names: Do not define custom methods like __my_custom_func__. The double underscore namespace is explicitly reserved by the Python core developers for future language enhancements.
3. 3
   Keep __new__ and __init__ Signatures Aligned: If __new__ takes arguments, __init__ should accept those exact arguments, as Python automatically routes instantiation parameters to both.
4. 4
   Prefer Context Managers Over __del__: Explicit resource termination via __enter__ and __exit__ avoids non-deterministic garbage collector behavior.

Multiple Choice Questions

1.

Which dunder method is the actual allocator responsible for creating and returning a new instance in memory before initialization? A. __init__ B. __new__ C. __call__ D. __prepare__

2.

What happens if you dynamically bind a dunder method to an instance dictionary, such as instance.__len__ = lambda: 10, and then execute len(instance)? A. len(instance) successfully returns 10. B. Python raises a TypeError stating that the object has no len(), because special methods are looked up on the class. C. Python updates the class dictionary dynamically and prints 10. D. A SyntaxError is raised immediately upon assignment.

3.

When subclassing an immutable built-in data type like str or int to modify its value prior to creation, which method must be overridden? A. __init__ B. __del__ C. __new__ D. __str__

4.

What is the expected return value of the __init__ method? A. The newly created instance (self) B. 0 for success or -1 for failure C. None D. A boolean True

5.

Why is relying on __del__ for releasing critical resources (like network connections or file descriptors) discouraged in production Python? A. __del__ is deprecated in Python 3. B. Garbage collection timing is non-deterministic, and cyclic references or abnormal interpreter exits can prevent __del__ from running promptly. C. __del__ cannot access instance attributes. D. Calling del obj always deletes the object instantly regardless of remaining references.