Python slots restrict which attributes an instance can have by defining __slots__ in a class. Instead of storing attributes in a flexible dictionary, instances with slots use fixed memory for only those attributes, reducing memory overhead and improving attribute access speed.

Use slots for large numbers of instances with few attributes, or when memory usage matters (data structures, game entities, ML datasets).

This example shows memory savings and access patterns with slots.

# run: python3 slots.py
# description: __slots__ for memory-efficient instances
 
import sys
 
class WithoutSlots:
    def __init__(self, x, y):
        self.x = x
        self.y = y
 
class WithSlots:
    __slots__ = ['x', 'y']
 
    def __init__(self, x, y):
        self.x = x
        self.y = y
 
# Create instances
no_slots = WithoutSlots(1, 2)
with_slots = WithSlots(1, 2)
 
# Compare memory usage
print(f"Without slots: {sys.getsizeof(no_slots.__dict__)} bytes for __dict__")
print(f"With slots: no __dict__ (fixed memory layout)")
 
print(f"Instance without slots: {sys.getsizeof(no_slots)} bytes")
print(f"Instance with slots: {sys.getsizeof(with_slots)} bytes")
 
# Try adding dynamic attributes
try:
    with_slots.z = 3
except AttributeError as e:
    print(f"Error: {e}")
 
# Slots instance is also faster to access
obj1 = WithoutSlots(10, 20)
obj2 = WithSlots(10, 20)
 
print(f"Without slots x: {obj1.x}")
print(f"With slots x: {obj2.x}")

Defining slots:

• __slots__ = ['attr1', 'attr2']: tuple or list of allowed attribute names
• No __dict__ created; attributes stored in fixed array
• Subclasses inherit parent's slots; can add more with their own __slots__

Special cases:

• Slot names can use descriptors for custom behavior
• __dict__ can be included in slots to allow dynamic attributes: __slots__ = ['x', '__dict__']
• __weakref__ in slots if weak references are needed

Trade-offs:

• Pro: ~40% memory savings per instance for typical objects
• Pro: Faster attribute access (direct array indexing vs dictionary lookup)
• Pro: Prevents typos (obj.lenght = 5 raises AttributeError instead of creating attribute)
• Con: Can't add attributes dynamically after class definition
• Con: Inheritance can get complex with slots in parent and child
• Con: Slots are class-level; instances can't have different slots

When to use:

• Large collections of objects (thousands or millions)
• Data structures with many instances but few attributes each
• Performance-critical inner loops
• Want to catch typos in attribute names

When to avoid:

• Small number of objects
• Need dynamic attributes
• Early in development (premature optimization)
• Simple scripts or one-off code