Behind the Underscores EP02: Object Initialization and Construction Methods (__new__, __init__, __del__)

In Python, object creation is controlled through a special set of methods: __new__, __init__, and __del__. These methods give us the flexibility to customize how objects are created, initialized, and destroyed. While __init__ is the most commonly used method, the other two, __new__ and __del__, allow for even more control, especially in advanced patterns such as singleton patterns, object pooling, and memory management.

In this article, we'll explore:

• What __new__, __init__, and __del__ are and how they work.
• Why and when to override them.
• Best practices and potential pitfalls.
• Real-world examples demonstrating how and when to use these methods.

Conceptual Overview

Before diving into the examples, it's crucial to understand the basic roles and behaviors of each method:

__new__(cls, *args, **kwargs)

• Responsible for: Object creation.
• What happens: The __new__ method is responsible for creating and returning a new instance of the class. It's called first, before __init__. If the object is immutable, __new__ has to return the fully initialized object, as the state of the object cannot be modified after creation.
• When to override: __new__ is usually overridden in advanced use cases, such as implementing design patterns like Singleton, Flyweight, subclassing immutable types like str, tuple, etc., or managing object creation efficiently like object pooling.
• Common use cases: Singleton pattern, Flyweight pattern, subclassing immutable types, object pooling (caching), metaclass manipulation.

__init__(self, *args, **kwargs)

• Responsible for: Initializing the object's state after creation.
• What happens: __init__ is called immediately after the object is created by __new__. Here, we initialize instance attributes and perform any necessary setup for the object.
• When to override: Most of the time, we override __init__ to initialize instance attributes, validate input, or perform some other setup. It’s the method that most developers interact with.
• Common use cases: Initializing object attributes, validating input, performing setup operations, dependency injection, and configuration parsing.

__del__(self)

• Responsible for: Cleaning up the object before it’s destroyed.
• What happens: The __del__ method is called when an object’s reference count drops to zero, indicating that the object is about to be garbage collected. It is intended for cleanup tasks, such as releasing external resources like file handles, network sockets.
• When to override: You override __del__ if you need to release external resources when the object is destroyed. However, it should be used with caution due to Python's non-deterministic garbage collection system.
• Common use cases: Closing files, releasing network connections, cleaning up database connections, and other resource management tasks.

Deep Dive into __new__

While most developers rarely need to override __new__, it provides powerful capabilities for advanced object creation strategies. Here are some scenarios where __new__ shines.

2.1 Immutable Subclasses

Python’s built-in immutable types like str, tuple, and frozenset cannot have their internal state changed once they are created. Therefore, any transformation on these types must happen at the moment of creation. That’s where __new__ comes into play.

Let’s look at that example:

class UpperStr(str):
    def __new__(cls, content):
        # Modify the content before the instance is created
        instance = super().__new__(cls, content.upper())
        return instance

s = UpperStr("hello")
print(s)  # Output: HELLO

In this example:

• The class UpperStr inherits from str which is immutable.
• The __new__ method is overridden to modify the string by converting it to uppercase before creating the object.
• Since str objects are immutable, the transformation must occur during creation, not after.

This approach ensures that once the object is created, it cannot be modified, preserving the immutability of the str type.

2.2 Singleton Pattern

The Singleton design pattern ensures that a class has only one instance throughout the application. To achieve this, we use __new__ to control object creation and ensure that only one instance is ever created.

class Singleton:
    _instance = None

    def __new__(cls):
        if cls._instance is None:
            cls._instance = super().__new__(cls)
        return cls._instance

In this example:

• The __new__ method checks whether the class has already created an instance.
• If the instance doesn't exist, it creates one using the super().__new__(cls) call.
• Subsequent attempts to create instances of the class will return the same object, enforcing the singleton behavior.

2.3 Object Pooling / Flyweight Pattern

The Flyweight pattern reduces memory usage by sharing objects that have the same state. Instead of creating a new instance every time, objects are reused from a shared pool.

class Flyweight:
    _cache = {}

    def __new__(cls, name):
        if name not in cls._cache:
            cls._cache[name] = super().__new__(cls)
        return cls._cache[name]

# Only one instance for each unique 'name'
obj1 = Flyweight("apple")
obj2 = Flyweight("apple")
obj3 = Flyweight("banana")

print(obj1 is obj2)  # Output: True (same instance)
print(obj1 is obj3)  # Output: False (different instance)

In this example:

• The Flyweight class caches instances in a dictionary based on the name passed to __new__.
• If an instance with the same name already exists, the existing instance is returned; otherwise, a new one is created.
• This reduces memory usage when dealing with a large number of objects with the same state.

Mastering __init__

The __init__ method is essential for initializing objects. While __new__ controls object creation, __init__ is where we initialize the object's state. Here are a few advanced use cases of __init__.

3.1 Dependency Injection

Dependency Injection (DI) is a technique where an object’s dependencies are provided (injected) to it during its initialization. This decouples the object from its dependencies, making it easier to test and modify.

class DatabaseConnection:
    def connect(self):
        return "Connected to the database."

class Service:
    def __init__(self, db_connection):
        self.db = db_connection

    def perform_task(self):
        return f"Task performed using {self.db.connect()}"

db = DatabaseConnection()
service = Service(db)

print(service.perform_task())  
# Output: Task performed using Connected to the database.

In this example:

• The Service class requires a DatabaseConnection object to function.
• Instead of creating the DatabaseConnection inside the Service class, it’s injected during initialization.
• This pattern makes the code more flexible and easier to test, as you can inject mock dependencies during unit tests.

3.2 Configuration Parsing

In many applications, we need to parse a configuration file or dictionary and initialize an object with the values. This can be easily handled by overriding __init__.

class Config:
    def __init__(self, config_dict):
        for k, v in config_dict.items():
            setattr(self, k, v)

config_dict = {"host": "localhost", "port": 8080}
config = Config(config_dict)

print(config.host)  # Output: localhost
print(config.port)  # Output: 8080

In this example:

• The Config class takes a dictionary and dynamically assigns each key-value pair as an attribute of the object using setattr().
• This pattern is useful when the attributes are not known ahead of time, such as in configuration management systems.

3.3 Runtime Type Checking

You can use __init__ to enforce type checks or validation on the input arguments. This is useful for ensuring that the object is always initialized with valid data.

class User:
    def __init__(self, age):
        if not isinstance(age, int):
            raise TypeError("Age must be an integer.")
        self.age = age

user = User(25)            # Works fine
user_invalid = User("25")  # Raises TypeError: Age must be an integer.

In this example:

• The User class ensures that the age argument passed to __init__ is an integer.
• If it's not an integer, a TypeError is raised, which prevents invalid data from being assigned to the object.

Subtleties of __del__

The __del__ method is Python’s destructor, called when an object is about to be garbage collected. However, it comes with caveats that should be carefully considered.

4.1 Basic Example

Here’s a simple example of using __del__ to clean up external resources such as file handles or network sockets:

class FileWriter:
    def __init__(self, path):
        self.file = open(path, 'w')

    def __del__(self):
        print("Closing file.")
        self.file.close()

In this example:

• The FileWriter class opens a file in __init__.
• The __del__ method ensures that the file is closed when the object is garbage collected.

4.2 Caveats and Best Practices

• Non-deterministic behavior: The garbage collector is non-deterministic, meaning that __del__ may not always be called when you expect.
• Reference cycles: If objects refer to each other in a cycle like object A refers to object B, and object B refers to object A, __del__ may not be called.
• Exceptions in __del__: If an exception is raised in __del__, it will be ignored, and you may not even be aware that there was an issue.

Use Context Managers Instead

For managing resources like file handles, it's better to use context managers (with statement) instead of relying on __del__.

class FileWriter:
    def __enter__(self):
        self.file = open("file.txt", "w")
        return self.file

    def __exit__(self, exc_type, exc_val, exc_tb):
        self.file.close()

with FileWriter() as file:
    file.write("Hello, world!")

In this example:

• The FileWriter class implements the context manager protocol (__enter__ and __exit__).
• The with statement ensures that the file is automatically closed when the block exits, even if an exception is raised.

Custom Metaclasses and __new__

Metaclasses are classes that define the behavior of other classes. __new__ in metaclasses is used to customize class creation itself.

class Meta(type):
    def __new__(cls, name, bases, dct):
        dct['created_by'] = "Meta"
        return super().__new__(cls, name, bases, dct)

class MyClass(metaclass=Meta):
    pass

obj = MyClass()
print(obj.created_by)  # Output: Meta

In this example:

• A custom metaclass Meta is defined, and its __new__ method adds a new attribute created_by to any class that uses it.
• MyClass uses Meta as its metaclass, and the created_by attribute is automatically added to the class.

When to Use What?

Use case	Method to override	Why?
Subclassing str, tuple, etc.	__new__	Immutable objects need early state setup
Enforcing Singleton/Flyweight	__new__	Control over object creation and reuse
Basic object state initialization	__init__	Assign values, validate input, inject deps
Releasing external resources	__del__ (rare)	Clean-up logic at end of object lifecycle
Safe resource management	__enter__/__exit__	Prefer over __del__ for determinism

By mastering these special methods, you gain precise control over object lifecycles, enable more efficient memory and resource usage, and lay the foundation for clean, extensible software architecture in Python.

The original post is here.