Weakref

Suppose you want to store as much data in memory as possible, but never want your program to use up all available memory. In Python we can use the weakref module for this purpose.

With weakref, and the WeakValueDictionary, we can add class instances to a dictionary without worrying about running out of memory. The values may be garbage collected at any time.

Example

This program uses the dataclass attribute for a simple class, and it adds instances of the class to either a WeakValueDictionary, or a regular dictionary.

Version 1 We create a WeakValueDictionary, and add 50 million instances of the StorageItem class to it.

Version 2 We create a regular Python dictionary, and add 50 million StorageItem instances to it as well.

Result The version that uses weakref can be garbage-collected, so it uses just 6.8 MB, while the dictionary causes memory usage of almost 8 GB.

import weakref
from dataclasses import dataclass

@dataclass
class StorageItem:
    name: str
    size: int

use_weak = True
print(use_weak)

if use_weak:
    # Version 1: use WeakValueDictionary to allow Python to garbage collect items on low memory.
    w = weakref.WeakValueDictionary()
    for i in range(50000000):
        w[i] = StorageItem("Parrot", i)
else:
    # Version 2: use dictionary, which uses memory that cannot be freed.
    d = {}
    for i in range(50000000):
        d[i] = StorageItem("Parrot", i)
Trueweakref memory:       6.8 MB
dictionary memory: 8000.0 MB

Summary

For caches, where we store data in memory for later use, weakref and WeakValueDictionary are ideal. We can add data to the cache without worrying about running out of memory.