Merge pull request #187 from BrianLusina/refactor/datastructures-lfu-cache

refactor(datastructures): lfu cache

Author: BrianLusina

DIRECTORY.md

@@ -549,6 +549,11 @@
     * [Test Hashmap](https://github.com/BrianLusina/PythonSnips/blob/master/datastructures/hashmap/test_hashmap.py)
   * Hashset
     * [Test My Hashset](https://github.com/BrianLusina/PythonSnips/blob/master/datastructures/hashset/test_my_hashset.py)
+  * Lfucache
+    * [Lfu Cache](https://github.com/BrianLusina/PythonSnips/blob/master/datastructures/lfucache/lfu_cache.py)
+    * [Lfu Cache Node](https://github.com/BrianLusina/PythonSnips/blob/master/datastructures/lfucache/lfu_cache_node.py)
+    * [Lfu Cache V2](https://github.com/BrianLusina/PythonSnips/blob/master/datastructures/lfucache/lfu_cache_v2.py)
+    * [Test Lfu Cache](https://github.com/BrianLusina/PythonSnips/blob/master/datastructures/lfucache/test_lfu_cache.py)
   * Linked Lists
     * Circular
       * [Circular Linked List Utils](https://github.com/BrianLusina/PythonSnips/blob/master/datastructures/linked_lists/circular/circular_linked_list_utils.py)
@@ -1154,7 +1159,6 @@
     * [Test Flatten Array](https://github.com/BrianLusina/PythonSnips/blob/master/tests/datastructures/test_flatten_array.py)
     * [Test Is Sorted How](https://github.com/BrianLusina/PythonSnips/blob/master/tests/datastructures/test_is_sorted_how.py)
     * [Test Length Of Missing Array](https://github.com/BrianLusina/PythonSnips/blob/master/tests/datastructures/test_length_of_missing_array.py)
-    * [Test Lfu Cache](https://github.com/BrianLusina/PythonSnips/blob/master/tests/datastructures/test_lfu_cache.py)
     * [Test List Ops](https://github.com/BrianLusina/PythonSnips/blob/master/tests/datastructures/test_list_ops.py)
     * [Test Manipulate Data](https://github.com/BrianLusina/PythonSnips/blob/master/tests/datastructures/test_manipulate_data.py)
     * [Test Min Max](https://github.com/BrianLusina/PythonSnips/blob/master/tests/datastructures/test_min_max.py)

datastructures/lfucache/README.md

@@ -17,4 +17,76 @@ To determine the least frequently used key, a use counter is maintained for each
 smallest use counter is the least frequently used key.
 
 When a key is first inserted into the cache, its use counter is set to 1 (due to the put operation). The use counter
-for a key in the cache is incremented either a get or put operation is called on it.
+for a key in the cache is incremented either a get or put operation is called on it.
+
+## Solution
+
+The LFU cache algorithm tracks how often each key is accessed to determine which keys to remove when the cache is full.
+It uses one hash map to store key-value pairs and another to group keys by their access frequency. Each group in this
+frequency hash map contains nodes arranged in a doubly linked list. Additionally, it keeps track of the current least
+frequency to quickly identify the least used keys. When the cache reaches its limit, the key with the lowest frequency
+is removed first, specifically from the head of the corresponding linked list.
+
+Each time a key is accessed, its frequency increases, and its position in the frequency hash map is updated, ensuring
+that the least used keys are prioritized for removal. This is where the doubly linked list is helpful, as the node being
+updated might be located somewhere in the middle of the list. Shifting the node to the next frequency level can be done
+in constant time, making the update process efficient.
+
+Let’s discuss the algorithm of the LFU cache data structure in detail. We maintain two hash maps, `lookup` and `frequencyMap`,
+and an integer, `minimum_frequency`, as follows:
+
+- `lookup` keeps the key-node pairs. 
+  - The node contains three values: `key`, `value`, and `frequency`.
+
+- `frequencyMap` maintains doubly linked lists against every frequency existing in the data. 
+  - For example, all the keys that have been accessed only once reside in the double linked list stored at `frequencyMap[1]`,
+    all the keys that have been accessed twice reside in the double linked list stored at `frequencyMap[2]`, and so on.
+
+- `minimum_frequency` keeps the frequency of the least frequently used key.
+
+Apart from the required functions i.e., Get and Put, we implement a helper function, `PromoteKey` that helps us maintain
+the order of the keys with respect to the frequency of their use. This function is implemented as follows:
+
+- First, retrieve the node associated with the key.
+- If node's `frequency` is 0, the key is new. We simply increment its `frequency` and insert it at the tail of the
+  linked list corresponding to the frequency 1
+- Otherwise, detach the `node` from its corresponding linked list. 
+  - If the corresponding linked list becomes empty after detaching the node, and the node’s `frequency` equals `minimum_frequency`, 
+    there's no key left with a frequency equal to `minimum_frequency`. Hence, increment `minimum_frequency`.
+- Increment `frequency` of the key
+- Insert node at the tail of the linked list associated with the frequency corresponding to the updated `frequency`.
+  - Before inserting it, check if the linked list exists. Suppose it doesn’t, create one.
+
+After implementing `PromoteKey()`, the LFU cache functions are implemented as follows:
+- `Get`: We check if the key exists in the cache. 
+  - If it doesn't, we return `None`
+  - Otherwise, we promote the key using `PromoteKey()` function and return the value associated with the key.
+- `Put`: We check if the key exists in the cache. 
+  - If it doesn't, we must add this (key, value) pair to our cache. 
+    - Before adding it, we check if the cache has already reached capacity. If it has, we remove the LFU key. To do that,
+      we remove the head node of the linked list associated with the frequency equal to `minimum_frequency`.
+    - Then we add the new key.
+  - If the key already exists, we simply update its value. 
+  - At the end of both cases, we adjust the frequency order of the key using `PromoteKey()`.
+
+![Solution 1](./images/solutions/lfu_cache_solution_1.png)
+![Solution 2](./images/solutions/lfu_cache_solution_2.png)
+![Solution 3](./images/solutions/lfu_cache_solution_3.png)
+![Solution 4](./images/solutions/lfu_cache_solution_4.png)
+![Solution 5](./images/solutions/lfu_cache_solution_5.png)
+![Solution 6](./images/solutions/lfu_cache_solution_6.png)
+![Solution 7](./images/solutions/lfu_cache_solution_7.png)
+![Solution 8](./images/solutions/lfu_cache_solution_8.png)
+![Solution 9](./images/solutions/lfu_cache_solution_9.png)
+![Solution 10](./images/solutions/lfu_cache_solution_10.png)
+
+### Time Complexity
+
+The time complexity of `PromoteKey()` is `O(1)` because the time taken to detach a node from a doubly linked list and
+insert a node at the tail of a linked list is `O(1)`. The time complexity of both Put and Get functions is `O(1)` because
+they utilize `PromoteKey()` and some other constant time operations.
+
+### Space Complexity
+
+The space complexity of this algorithm is linear, `O(n)`, where `n` refers to the capacity of the data structure. This
+is the space occupied by the hash maps.

datastructures/lfucache/__init__.py

@@ -1,127 +1,5 @@
-from collections import defaultdict
-from typing import Any, Union, Dict
+from datastructures.lfucache.lfu_cache_node import LfuCacheNode
+from datastructures.lfucache.lfu_cache import LFUCache
+from datastructures.lfucache.lfu_cache_v2 import LFUCacheV2
 
-from datastructures.linked_lists.doubly_linked_list import DoublyLinkedList
-from datastructures.linked_lists.doubly_linked_list.node import DoubleNode
-
-
-class LfuCacheNode(DoubleNode):
-    def __init__(self, data):
-        super().__init__(data)
-        self.frequency = 1
-
-
-class LFUCache:
-    def __init__(self, capacity: int):
-        """
-        Initializes an instance of a LFUCache
-        @param capacity: Capacity of the cache
-        @type capacity int
-
-        1. Dict named node self._lookup for retrieval of all nodes given a key. O(1) time to retrieve a node given a key
-        2. Each frequency has a DoublyLinkedList stored in self._frequency where key is the frequency and value is an
-        object of DoublyLinkedList
-        3. minimum frequency through all nodes, this can be maintained in O(1) time, taking advantage of the fact that
-        the frequency can only increment by 1. use the following 2 rules:
-            i. Whenever we see the size of the DoublyLinkedList of current min frequency is 0, increment min_frequency
-                by 1
-            ii. Whenever we put in a new (key, value), the min frequency must be 1 (the new node)
-        """
-        self.capacity = capacity
-        self._current_size = 0
-        self._lookup = dict()
-        self._frequency: Dict[int, DoublyLinkedList] = defaultdict(DoublyLinkedList)
-        self._minimum_frequency = 0
-
-    def __update(self, node: LfuCacheNode):
-        """
-        Helper function used in 2 cases:
-            1. When get(key) is called
-            2. When put(key, value) is called and key exists
-
-        Common point of the 2 cases:
-            1. no new node comes in
-            2. node is visited one more time -> node.frequency changed -> thus the place of this node will change
-
-        Logic:
-            1. Pop node from 'old' DoublyLinkedList with frequency
-            2. Append node to 'new' DoublyLinkedList with frequency + 1
-            3. If 'old' DoublyLinkedList has size 0 & self.minimum_frequency is frequency, update self.minimum_frequency
-                to frequency + 1
-
-        Complexity Analysis:
-            Time Complexity: O(1) time
-
-        @param node: Node to update in the Cache
-        @type node LfuCacheNode
-        """
-        frequency = node.frequency
-
-        # pop the node from the 'old' DoublyLinkedList
-        self._frequency[frequency].delete_node(node)
-
-        if self._minimum_frequency == frequency and not self._frequency[frequency]:
-            self._minimum_frequency += 1
-
-        node.frequency += 1
-        frequency = node.frequency
-
-        # add to 'new' DoublyLinkedList with new frequency
-        self._frequency[frequency].prepend(node)
-
-    def get(self, key: int) -> Union[Any, None]:
-        """
-        Gets an item from the Cache given the key
-        @param key: Key to use to fetch data from Cache
-        @return: Data mapped to the key
-        """
-        if key not in self._lookup:
-            return None
-
-        node = self._lookup[key]
-        data = node.data
-        self.__update(node)
-        return data
-
-    def put(self, key: int, value: Any) -> None:
-        """
-        If key is already present in the self._lookup, we perform same operations as get, except updating the node data
-        to new value
-
-        Otherwise, below operations are performed:
-        1. If cache reaches capacity, pop least frequently used item.
-            2 Facts:
-            a. we maintain self._minimum_frequency, minimum possible frequency in cache
-            b. All cache with the same frequency are stored as a DoublyLinkedList, with recently used order (Always
-                append to head).
-
-            Consequence is that the tail of the DoublyLinkedList with self._minimum_frequency is the least recently used
-            one, pop it.
-
-        2. Add new node to self._lookup
-        3. add new node to DoublyLinkedList with frequency of 1
-        4. reset minimum_frequency to 1
-
-        @param key: Key to use for lookup
-        @param value: Value to store in the cache
-        @return: None
-        """
-
-        if self.capacity == 0:
-            return None
-
-        if key in self._lookup:
-            node = self._lookup[key]
-            self.__update(node)
-            node.data = value
-        else:
-            if self._current_size == self.capacity:
-                node = self._frequency[self._minimum_frequency].pop()
-                self._lookup.pop(node.key)
-                self._current_size -= 1
-
-            node = DoubleNode(data=value, key=key)
-            self._lookup[key] = node
-            self._frequency[1].append(node)
-            self._minimum_frequency = 1
-            self._current_size += 1
+__all__ = ["LFUCache", "LFUCacheV2", "LfuCacheNode"]