Memory Management: Reference Counting, Cyclic GC & Context Managers (with)

1Concept

CPython uses Reference Counting (ob_refcnt) as primary GC: objects are deallocated the instant refcount drops to 0. A 3-generation cyclic garbage collector resolves circular reference graphs. Context managers (with open() as f:) guarantee cleanup via __enter__ and __exit__.

2Architecture Diagram

PyObject Structure:
 +-------------------------------------+
 | ob_refcnt : Number of active ptrs   | ──► When refcnt == 0: Memory immediately freed!
 | ob_type   : Pointer to type object  |
 | Data payload                        |
 +-------------------------------------+

3Code Example

import sys

class CustomResource:
    def __init__(self, name):
        self.name = name

    def __enter__(self):
        print(f"[Context Manager] Acquired {self.name}")
        return self

    def __exit__(self, exc_type, exc_val, exc_tb):
        print(f"[Context Manager] Released {self.name}")
        return False # Propagate exceptions if any

with CustomResource("Database Connection") as res:
    print(f"Working with {res.name}...")

4Expected Output

[Context Manager] Acquired Database Connection
Working with Database Connection...
[Context Manager] Released Database Connection

5Key Takeaways

• ✓Use context managers (with statement) to guarantee file/socket closure.
• ✓Reference counting provides deterministic deallocation except for reference cycles.
• ✓Use weakref to break cyclic references in object graphs.