Python
functools
Learn caching with functools.lru_cache.
functools and lru_cache
The functools module provides helpers for higher-order functions. The most common is lru_cache, which keeps results of repeated calls in memory (memoization).
For example, recursive fib(n) is expensive; each call makes two sub-calls. Decorating it with @lru_cache(maxsize=None) stores results in a dict; if the same n shows up again it returns instantly.
from functools import lru_cache
@lru_cache(maxsize=None)
def fib(n):
if n < 2:
return n
return fib(n-1) + fib(n-2)
maxsize=None means an unlimited cache; a small maxsize evicts the least recently used entries (LRU).
Cache behavior and limits
lru_cache behaves like a dict: the key is the arguments passed to the function. If the same arguments come back, no computation happens.
Keeping maxsize small controls memory use. With maxsize=128, only the last 128 distinct calls are kept; a new call evicts the oldest. fib.cache_info() shows how many hits and misses occurred.
Note: arguments to a cached function must be hashable. You cannot pass a list or dict; use an immutable tuple instead.
Exercise Cache fib with lru_cache
Decorate the recursive fib(n) with @lru_cache(maxsize=None). The base case n < 2 returns n; otherwise it returns fib(n-1) + fib(n-2). End with print(fib(10)).
Show example solution
from functools import lru_cache
@lru_cache(maxsize=None)
def fib(n):
if n < 2:
return n
return fib(n-1) + fib(n-2)
print(fib(10))
Test yourself
-
Match the lru_cache concepts
- lru_cache
- Result cache
- maxsize=None
- Unlimited cache
- maxsize=128
- LRU eviction
- fib(n-1)+fib(n-2)
- Recursive step
-
Add the lru_cache decorator
from functools import lru_cache
@____(maxsize=None)
def fib(n):
if n < 2:
return n
return fib(n-1) + fib(n-2)
Show answer
from functools import lru_cache
@lru_cache(maxsize=None)
def fib(n):
if n < 2:
return n
return fib(n-1) + fib(n-2)
-
Which argument type CANNOT be passed to an lru_cache function?
Show answer
List — A list is not hashable; it cannot be a cache key. Use a tuple instead.
Next lesson: functools.partial
