Serialization Management

10.1 It's My Way or the Highway!

Sometimes your objects hold a lot of references to various utility objects that you don't want to pass over the network, or that can't be passed over the network: references to files, databases, etc.

To make serialization work in such cases, they came up with the idea of giving a class the ability to manage its own serialization. For this, special methods are used: __reduce__(), __getstate__(), __setstate__(). These methods allow you to specify how objects should be serialized and restored.

Main methods of controlled serialization:

• __reduce__(): Specifies how an object should be serialized.
• __getstate__(): Returns the state of the object for serialization.
• __setstate__(self, state): Restores the object from the state.

I'll explain them in more detail below, and how to use them together.

10.2 The __reduce__() Method

The __reduce__() method returns a tuple that specifies how an object should be serialized and deserialized. The tuple normally contains:

• A reference to a function or class that will be used to restore the object.
• A tuple of arguments for this function or class.
• Additional object state (if necessary).

Example:

import pickle

class CustomClass:
    def __init__(self, value):
        self.value = value
            
    def __reduce__(self):
        return (self.__class__, (self.value,))
            
    def __repr__(self):
        return f"CustomClass(value={self.value})"
            
# Create an object
obj = CustomClass(42)
            
# Serialize the object
serialized_obj = pickle.dumps(obj)
print("Serialized object:", serialized_obj)
            
# Deserialize the object
deserialized_obj = pickle.loads(serialized_obj)
print("Deserialized object:", deserialized_obj)

By default, the __reduce__() function behaves like this:

class CustomClass:
    def __init__(self, value):
        self.value = value
        
    def __reduce__(self):
        # Define the class
        cls = self.__class__
        # Constructor arguments
        args = (self.value,)
        # Object state
        state = self.__dict__
        return (cls, args, state)

It returns a tuple consisting of three objects:

• Reference to the current class
• Constructor arguments (tuple)
• Reference to the current state of the object

If you're cool with this behavior — you can skip overriding __reduce__().

10.3 Reading and Writing State

The __getstate__() and __setstate__() Methods

These methods are used to manage the state of an object during serialization and deserialization.

• __getstate__(): Returns the state of the object that should be serialized.
• __setstate__(self, state): Restores the object from the state.

Example:

Let's say we want to save not all fields of an object but exclude some of them. For this, in the __getstate__() method, you need to:

1. Copy the current state of the object (set by the utility field __dict__) into a separate variable — a dictionary state.
2. Remove all fields from it that don't need to be serialized.
3. Return the resulting object as the result of the __getstate__() function.

import pickle

class CustomClass:
    def __init__(self, value):
        self.value = value
        self.internal_state = "internal"
            
    def __getstate__(self):
        state = self.__dict__.copy()
        del state['internal_state']  # Exclude the internal state
        return state
            
    def __setstate__(self, state):
        self.__dict__.update(state)
        self.internal_state = "restored internal"  # Restore internal state
            
    def __repr__(self):
        return f"CustomClass(value={self.value}, internal_state={self.internal_state})"
            
# Create an object
obj = CustomClass(42)
print("Original object:", obj)
            
# Serialize the object
serialized_obj = pickle.dumps(obj)
print("Serialized object:", serialized_obj)
            
# Deserialize the object
deserialized_obj = pickle.loads(serialized_obj)
print("Deserialized object:", deserialized_obj)

During deserialization, in the __setstate__() function, we do two things:

1. Update the current state of the object using the update() method.
2. The internal_state field (and other unserializable fields) get new values.

2

Task

Python SELF EN, level 22, lesson 2

Locked

Using the reduce() method

Using the reduce() method

Open

2

Task

Python SELF EN, level 22, lesson 2

Locked

Excluding Non-serializable Fields

Excluding Non-serializable Fields

Open

1

Survey/quiz

Working with Directories, level 22, lesson 2

Unavailable

Working with Directories

Working with Directories

Open