OTOH that’s a lot less useful now that functools.lru_cache exists: it’s more specialised so it’s lighter, more efficient, and thread-safe. So unless you have ex

OTOH that’s a lot less useful now that functools.lru_cache exists: it’s more specialised so it’s lighter, more efficient, and thread-safe. So unless you have extended flexibility requirements around your LRU, OD loses a lot there. And if you’re using a FIFO cache, threading a regular dict through a separate fifo (whether linked list or deque) is more efficient in my experience of implementing both S3 and Sieve.