Memory Management in Python - The Basics

Speaker

Nina Zakharenko

How does Python store object in memory
- name -> reference -> object
Python Object Types
- Simple (e.g., numbers, strings)
- Containers (dict, list, user defined-class)
Every Python object holds
- type
- value
- ref count
Reference Count
- Decrease ref count
  - del
    - remove the name as a reference instead of delete the object
    - decrease the reference count by 1
  - going out of scope
Global Namespace
- never go out of scope → refcount never reach 0
- Avoid putting large or complex objects in the global namespace
Common Garbage Collection Techniques
- Reference Counting
  - Concept
    - add / remove ref
    - delete object if refcount reaches 0
  - The good
    - Easy to implement
    - When refcount is 0 objects are immediately delete.
  - The bad
    - space overhead
    - execution overhead
  - The ugly
    - not thread safe
    - doesn't detect cyclical ref
- Tracing (mark and sweep)
What algorithm does Python use?
- Reference Counting + Generational
__slot__
- Attribute Declarations
- When to use
  - If we're going to create lots of instances
  - If we know the properties in advance