Implement LRU cache with using @functools.lru_decorator in python

Question

So I've been trying to implement an LRU cache for my project, Using the python functools lru_cache. As a reference I used this. The following is the code is used from the reference.

def timed_lru_cache(maxsize, seconds):
    def wrapper_cache(func):
        func = lru_cache(maxsize=maxsize)(func)
        func.lifetime = timedelta(seconds=seconds)
        func.expiration = datetime.utcnow() + func.lifetime

        @wraps(func)
        def wrapped_func(*args, **kwargs):
            if datetime.utcnow() >= func.expiration:
                func.cache_clear()
                func.expiration = datetime.utcnow() + func.lifetime

            return func(*args, **kwargs)

        return wrapped_func

    return wrapper_cache

    @timed_lru_cache(maxsize=config.cache_size, seconds=config.ttl)
    def load_into_cache(id):
        return object

In the wrapped func part, the func.cache_clear(), clears the entire cache along with all the items. I need help to remove only elements past its expiretime after inserting. Is there any work around?

Answer 1

I don't think it's so easy to adapt the existing lru_cache, and I don't think that linked method is very clear.

Instead I implemented a timed lru cache from scratch. See the docstring at the top for usage.

It stores a key based on the args and kwargs of the inputs, and manages two structures:

• A mapping of key => (expiry, result)
• A list of recently used, where the first item is the least recently used

Every time you try to get an item, the key is looked up in the "recently used" list. If it isn't there, it gets added to the list and the mapping. If it is there, we check if the expiry is in the past. If it is, we recalculate the result, and update. Otherwise we can just return whatever is in the mapping.

from datetime import datetime, timedelta
from functools import wraps
from typing import Any, Dict, List, Optional, Tuple


class TimedLRUCache:
    """ Cache that caches results based on an expiry time, and on least recently used.
    
        Items are eliminated first if they expire, and then if too many "recent" items are being
        stored. 
        
        There are two methods of using this cache, either the `get` method`, or calling this as a
        decorator. The `get` method accepts any arbitrary function, but on the parameters are
        considered in the key, so it is advisable not to mix function.
        
        >>> cache = TimedLRUCache(5)
        >>> def foo(i):
        ...     return i + 1
        
        >>> cache.get(foo, 1)  # runs foo
        >>> cache.get(foo, 1)  # returns the previously calculated result
        
        As a decorator is more familiar:
        
        >>> @TimedLRUCache(5)
        ... def foo(i):
        ...     return i + 1
        
        >>> foo(1)  # runs foo
        >>> foo(1)  # returns the previously calculated result
        
        
        Either method can allow for fine-grained control of the cache:
        
        >>> five_second_cache = TimedLRUCache(5)
        >>> @five_second_cache
        ... def foo(i):
        ...     return i + 1
        
        >>> five_second_cache.clear_cache()  # resets the cache (clear every item)
        >>> five_second_cache.prune()  # clear invalid items
    """
    _items: Dict[int, Tuple[datetime, Any]]
    _recently_added: List[int]

    delta: timedelta
    max_size: int

    def __init__(self, seconds: Optional[int] = None, max_size: Optional[int] = None):
        self.delta = timedelta(seconds=seconds) if seconds else None
        self.max_size = max_size

        self._items = {}
        self._recently_added = []
        
    def __call__(self, func):
        @wraps(func)
        def wrapper(*args, **kwargs):
            return self.get(func, args, kwargs)
        return wrapper

    @staticmethod
    def _get_key(args, kwargs) -> int:
        """ Get the thing we're going to use to lookup items in the cache. """
        key = (args, tuple(sorted(kwargs.items())))
        return hash(key)

    def _update(self, key: int, item: Any) -> None:
        """ Make sure an item is up to date. """
        if key in self._recently_added:
            self._recently_added.remove(key)
        # the first item in the list is the least recently used
        self._recently_added.append(key)
        self._items[key] = (datetime.now() + self.delta, item)

        # when this function is called, something has changed, so we can also sort out the cache
        self.prune()

    def prune(self):
        """ Clear out everything that no longer belongs in the cache

            First delete everything that has expired. Then delete everything that isn't recent (only
            if there is a `max_size`).
        """
        # clear out anything that no longer belongs in the cache.
        current_time = datetime.now()
        # first get rid of things which have expired
        for key, (expiry, item) in self._items.items():
            if expiry < current_time:
                del self._items[key]
                self._recently_added.remove(key)
        # then make sure there aren't too many recent items
        if self.max_size:
            self._recently_added[:-self.max_size] = []

    def clear_cache(self):
        """ Clear everything from the cache """
        self._items = {}
        self._recently_added = []

    def get(self, func, args, kwargs):
        """ Given a function and its arguments, get the result using the cache

            Get the key from the arguments of the function. If the key is in the cache, and the
            expiry time of that key hasn't passed, return the result from the cache.

            If the key *has* expired, or there are too many "recent" items, recalculate the result,
            add it to the cache, and then return the result.
        """
        key = self._get_key(args, kwargs)
        current_time = datetime.now()
        if key in self._recently_added:
            # there is something in the cache
            expiry, item = self._items.get(key)
            if expiry < current_time:
                # the item has expired, so we need to get the new value
                new_item = func(*args, **kwargs)
                self._update(key, new_item)
                return new_item
            else:
                # we can use the existing value
                return item
        else:
            # never seen this before, so add it
            new_item = func(*args, **kwargs)
            self._update(key, new_item)
            return new_item