[Medium] 146. LRU Cache
[Medium] 146. LRU Cache
Design a data structure that follows the constraints of a Least Recently Used (LRU) cache.
Implement the LRUCache class:
LRUCache(int capacity)Initialize the LRU cache with positive sizecapacity.int get(int key)Return the value of thekeyif the key exists, otherwise return-1.void put(int key, int value)Update the value of thekeyif thekeyexists. Otherwise, add thekey-valuepair to the cache. If the number of keys exceeds thecapacityfrom this operation, evict the least recently used key.
The functions get and put must each run in O(1) average time complexity.
Examples
Example 1:
Input
["LRUCache", "put", "put", "get", "put", "get", "put", "get", "get", "get"]
[[2], [1, 1], [2, 2], [1], [3, 3], [2], [4, 4], [1], [3], [4]]
Output
[null, null, null, 1, null, -1, null, -1, 3, 4]
Explanation
LRUCache lRUCache = new LRUCache(2);
lRUCache.put(1, 1); // cache is {1=1}
lRUCache.put(2, 2); // cache is {1=1, 2=2}
lRUCache.get(1); // return 1
lRUCache.put(3, 3); // LRU key was 2, evicts key 2, cache is {1=1, 3=3}
lRUCache.get(2); // returns -1 (not found)
lRUCache.put(4, 4); // LRU key was 1, evicts key 1, cache is {4=4, 3=3}
lRUCache.get(1); // return -1 (not found)
lRUCache.get(3); // return 3
lRUCache.get(4); // return 4
Constraints
1 <= capacity <= 30000 <= key <= 10^40 <= value <= 10^5- At most
2 * 10^5calls will be made togetandput.
Clarification Questions
Before diving into the solution, here are 5 important clarifications and assumptions to discuss during an interview:
-
LRU definition: What does “Least Recently Used” mean? (Assumption: The item that hasn’t been accessed for the longest time - need to track access order)
-
Cache operations: What operations should the cache support? (Assumption: get(key) - retrieve value, put(key, value) - insert/update, both operations mark item as recently used)
-
Eviction policy: When should we evict items? (Assumption: When cache is full and we need to add a new item, evict the least recently used item)
-
Capacity: What is the cache capacity? (Assumption: Fixed capacity specified in constructor - cannot exceed this limit)
-
Return values: What should get() return if key doesn’t exist? (Assumption: Return -1 - key not found in cache)
Interview Deduction Process (20 minutes)
Step 1: Brute-Force Approach (5 minutes)
Use a hash map to store key-value pairs and a list/array to track access order. For get, search the list to find the key, move it to the end, and return the value. For put, add/update the hash map and move the key to the end of the list. When capacity is exceeded, remove the first element from the list. This approach has O(n) time for get operations due to list searching, which doesn’t meet the O(1) requirement.
Step 2: Semi-Optimized Approach (7 minutes)
Use a hash map and a doubly linked list: hash map stores key-to-node mappings, doubly linked list maintains access order. For get, use hash map to find the node in O(1), move it to the end of the list. For put, add/update hash map, add node to end of list, remove head if capacity exceeded. However, implementing a doubly linked list from scratch requires careful pointer management and can be error-prone.
Step 3: Optimized Solution (8 minutes)
Use hash map + std::list: hash map stores key -> iterator mappings, std::list maintains access order with pairs (key, value). For get, use hash map to get iterator in O(1), use splice to move node to end in O(1). For put, if key exists, update value and move to end. If new key, add to end and remove head if capacity exceeded. This achieves O(1) for both operations using standard library containers. The key insight is that std::list::splice allows O(1) node movement, and hash map provides O(1) lookup, combining to achieve O(1) operations.
Solution: Hash Map + Doubly Linked List
Time Complexity: O(1) for both get and put
Space Complexity: O(capacity)
We use a combination of hash map and doubly linked list to achieve O(1) operations. The hash map stores key-to-iterator mappings, and the doubly linked list maintains the order of recently used items.
Solution: Using std::list with splice
#include <list>
#include <unordered_map>
class LRUCache {
public:
LRUCache(int capacity) : capacity_(capacity) {
}
int get(int key) {
auto it = hashMap_.find(key);
if(it != hashMap_.end()) {
keyList_.splice(keyList_.end(), keyList_, it->second.second);
return it->second.first;
}
return -1;
}
void put(int key, int value) {
if(get(key) != -1) {
hashMap_[key].first = value;
return;
}
if(hashMap_.size() >= capacity_) {
int removeKey = keyList_.front();
keyList_.pop_front();
hashMap_.erase(removeKey);
}
keyList_.push_back(key);
hashMap_[key] = {value, --keyList_.end()};
}
private:
int capacity_;
list<int> keyList_;
unordered_map<int, pair<int, list<int>::iterator>> hashMap_;
};
/**
* Your LRUCache object will be instantiated and called as such:
* LRUCache* obj = new LRUCache(capacity);
* int param_1 = obj->get(key);
* obj->put(key,value);
*/
Key Points
list<int> keyList_: Maintains the order of keys, with front being LRU and back being MRUunordered_map<int, pair<int, list<int>::iterator>>: Maps key to (value, iterator) pairlist::splice(): O(1) operation to move nodes without copyingget()operation: Moves accessed key to end (most recently used) usingsplice()put()operation:- If key exists (via
get()), updates value and returns early - If capacity is full, removes front (LRU) before inserting
- Always inserts new key at end (most recently used)
- If key exists (via
Thread-Safe LRU Cache
Thread-safe version using mutex for concurrent access.
#include <list>
#include <unordered_map>
#include <mutex>
#include <shared_mutex>
class ThreadSafeLRUCache {
public:
ThreadSafeLRUCache(int capacity) : capacity_(capacity) {
}
int get(int key) {
unique_lock<shared_mutex> lock(mtx_); // Exclusive lock for read+modify
auto it = hashMap_.find(key);
if(it != hashMap_.end()) {
keyList_.splice(keyList_.end(), keyList_, it->second.second);
return it->second.first;
}
return -1;
}
void put(int key, int value) {
unique_lock<shared_mutex> lock(mtx_); // Exclusive lock for write
auto it = hashMap_.find(key);
if(it != hashMap_.end()) {
// Key exists, update value and move to end
hashMap_[key].first = value;
keyList_.splice(keyList_.end(), keyList_, it->second.second);
return;
}
if(hashMap_.size() >= capacity_) {
int removeKey = keyList_.front();
keyList_.pop_front();
hashMap_.erase(removeKey);
}
keyList_.push_back(key);
hashMap_[key] = {value, --keyList_.end()};
}
size_t size() const {
shared_lock<shared_mutex> lock(mtx_);
return hashMap_.size();
}
private:
int capacity_;
list<int> keyList_;
unordered_map<int, pair<int, list<int>::iterator>> hashMap_;
mutable shared_mutex mtx_; // Use shared_mutex for read-write lock
};
// Example usage:
// ThreadSafeLRUCache cache(2);
// cache.put(1, 1);
// cache.put(2, 2);
// int val = cache.get(1); // returns 1
// cache.put(3, 3); // evicts key 2
Thread-Safe Implementation Details
shared_mutex: Allows multiple concurrent reads when no writes are happeningunique_lock<shared_mutex>: Exclusive lock for bothget()andput()since they modify the listshared_lock<shared_mutex>: Shared lock for read-only operations likesize()- Note: The thread-safe
put()callsget()which requires careful lock management - both methods use exclusive locks
How the Algorithm Works
Data Structure Design
Hash Map: Doubly Linked List:
key -> (value, iterator) [1] <-> [2] <-> [3]
(LRU) (MRU)
Operation Flow
Get Operation:
- Look up key in hash map → O(1)
- If found, move node to end (most recently used) using
splice()→ O(1) - Return value
Put Operation:
- Check if key exists via
get()→ O(1) - If exists: update value (already moved to end by
get()) → O(1) - If new:
- Check capacity
- If full: remove front node (LRU) → O(1)
- Insert at end → O(1)
Example Walkthrough
capacity = 2
put(1, 1): cache = {1: (1, it1)}
list: [1]
put(2, 2): cache = {1: (1, it1), 2: (2, it2)}
list: [1] <-> [2]
get(1): Move 1 to end using splice
list: [2] <-> [1]
return 1
put(3, 3): get(3) returns -1, capacity full
Remove front (key 2), insert 3 at end
cache = {1: (1, it1), 3: (3, it3)}
list: [1] <-> [3]
Complexity Analysis
| Operation | Time | Space |
|---|---|---|
get(key) |
O(1) | O(1) |
put(key, value) |
O(1) | O(1) |
| Overall | O(1) | O(capacity) |
Why This Approach Works
- Hash map: Provides O(1) lookup by key
- Doubly linked list: Maintains insertion order and allows O(1) removal/insertion
- Iterator storage: Hash map stores iterators to list nodes for O(1) access
splice()optimization: Moves nodes without copying, maintaining O(1) complexity
Edge Cases
- Capacity = 1: Only one item can exist
- Get non-existent key: Returns -1
- Update existing key: Moves to end, doesn’t increase size
- Multiple puts: Evicts oldest when capacity exceeded
Common Mistakes
- Not moving to end on get: Must update access order
- Wrong eviction order: Remove from front (LRU), not back
- Invalid iterator after modification: Use
splice()which preserves iterators - Not handling existing key in put: Must check and update, not just insert
Related Problems
- 460. LFU Cache - Least Frequently Used
- 432. All O`one Data Structure
- 588. Design In-Memory File System