Why?

There are only two hard things in Computer Science: cache invalidation and naming things.

– Phil Karlton

So, caches are fairly easy, but their invalidation is kind of hard to get right.

What type of cache is used?

The default implementation is the lru_cache of the stdlib but you can drop in your own.

How?

Python features garbage collection aka. automatic memory management. Let’s make use of it:

from xcache import cache_gen

request_cache = cache_gen(lambda: request)  # default cache_impl=lru_cache

Here, we defined our own cache that needs the global request for invalidating the cache. As soon as request is garbage-collected, the cached data is freed.

@request_cache()
def fib(n):
    return fib(n-1) + fib(n-2) if n > 1 else 1

Here, we see how to decorate a function to work like a lru_cache.

Now, let’s see how this works:

class Request(): pass

request = Request()

print(fib(10))
print(fib(20))
print(fib.cache_info())  # fib cache contains 2 items

request = Request()      # invalidates all caches

print(fib.cache_info())  # fib cache contains 0 items
print(fib(10))
print(fib(20))

Context Manager for more more control over cache invalidation

If you need more control, the context manager clean_caches is what you need:

from xcache import cached, clean_caches

@cached()
def fib(n):
    return fib(n-1) + fib(n-2) if n > 1 else 1

with clean_caches():
    print(fib(10))
    print(fib(20))
    print(fib.cache_info())  # fib cache contains 2 items
print(fib.cache_info())      # fib cache contains 0 items

You can even specify what object the caches should be attached to:

@cached()
def fib(n):
    return fib(n-1) + fib(n-2) if n > 1 else 1

with clean_caches(Request()) as request:
    print(fib(10))
    print(fib(20))
    print(fib.cache_info())  # fib cache contains 2 items
print(fib.cache_info())      # fib cache contains 0 items

Can these context managers be nested?

Sure. At each entrance and and exit of each context, all associated caches are emptied.

Conclusion