Least Recently Used (LRU) cache algorithm

This is an older version aimed for older JS environments lacking some post-ES5 features like Map and Symbol. If you're targeting a modern web browser or Nodejs, consider using the latest version instead.

A finite key-value cache using the Least Recently Used (LRU) cache algorithm where the most recently used objects are keept in cache while less recently used items are purged.

This implementation is compatible with most JavaScript environments (including ye olde browser) and is very efficient.

Terminology & design

• Based on a doubly-linked list for low complexity random shuffling of entries.

• The cache object iself has a "head" (least recently used entry) and a "tail" (most recently used entry).

• The "head" and "tail" are "entries" -- an entry might have a "newer" and an "older" entry (doubly-linked, "older" being close to "head" and "newer" being closer to "tail").

• Key lookup is done through a key-entry mapping native object, which on most platforms mean O(1) complexity. This comes at a very low memory cost (for storing two extra pointers for each entry).

Fancy ASCII art illustration of the general design:

           entry             entry             entry             entry        
           ______            ______            ______            ______       
          | head |.newer => |      |.newer => |      |.newer => | tail |      
.oldest = |  A   |          |  B   |          |  C   |          |  D   | = .newest
          |______| <= older.|______| <= older.|______| <= older.|______|      
                                                                             
       removed  <--  <--  <--  <--  <--  <--  <--  <--  <--  <--  <--  added

Example

let c = new LRUCache(3)
c.put('adam',   29)
c.put('john',   26)
c.put('angela', 24)
c.toString()        // -> "adam:29 < john:26 < angela:24"
c.get('john')       // -> 26

// Now 'john' is the most recently used entry, since we just requested it
c.toString()        // -> "adam:29 < angela:24 < john:26"
c.put('zorro', 141) // -> {key:adam, value:29}

// Because we only have room for 3 entries, put-ing 'zorro' purged 'adam'
c.toString()        // -> "angela:24 < john:26 < zorro:141"

Usage

Just copy the code on lru.js — for minimal functionality, you only need the lines up until the comment that says "Following code is optional".

If you're really into package managers and love having lots of complicated little files in your project, you can use npm install lru-fast

Additionally:

• Run tests with npm test
• Run benchmarks with npm run benchmark

API

// An entry holds the key and value, and pointers to any older and newer entries.
interface Entry<K,V> {
  older :Entry<K,V>;
  newer :Entry<K,V>;
  key   :K;
  value :V;
}

export class LRUCache<K,V> {
  // Construct a new cache object which will hold up to limit entries.
  // When the size == limit, a `put` operation will evict the oldest entry.
  constructor(limit :number);

  // Current number of items
  size: number;

  // Maximum number of items this map can hold
  limit: number;

  // Least recently-used entry
  oldest: Entry<K,V>;

  // Most recently-used entry
  newest: Entry<K,V>;

  // Put <value> into the cache associated with <key>. Replaces any existing entry
  // with the same key.
  // Returns any entry which was removed to make room for a new entry, or undefined.
  // Note: The behavior of this method changed between v0.1 and v0.2 where in v0.1
  // putting multiple values with the same key would store all values (accessible
  // via forEach and other forms of traversal.) v0.2 stores exactly one value per key.
  put(key :K, value :V) : Entry<K,V> | undefined;

  // Remove the least recently-used (oldest) entry from the cache.
  // Returns the removed entry, or undefined if the cache was empty.
  shift() : Entry<K,V> | undefined;

  // Get and register recent use of <key>.
  // Returns the value associated with <key> or undefined if not in cache.
  get(key :K) : V | undefined;

  // Check if <key> is in the cache without registering recent use. Feasible if
  // you do not want to chage the state of the cache, but only "peek" at it.
  // Returns the entry associated with <key> if found, or undefined if not found.
  //
  // Note: The entry returned is managed by the cache (until purged) and thus
  // contains members with strong references which might be altered at any time by
  // the cache object. You should look at the returned entry as being immutable.
  find(key :K) : V | undefined;

  // Update the value of entry with <key>.
  // Returns the old value, or undefined if entry was not in the cache.
  set(key :K, value :V) : V | undefined;

  // Remove entry <key> from cache and return its value.
  // Returns the removed value, or undefined if not found.
  remove(key :K) : V | undefined;

  // Removes all entries
  removeAll();

  // Return an array containing all keys of entries stored in the cache object, in
  // arbitrary order.
  keys() : Array<K>;

  // Call `fun` for each entry. Starting with the newest entry if `desc` is a true
  // value, otherwise starts with the oldest (head) enrty and moves towards the tail.
  // context, Object key, Object value, LRUCache self
  forEach(
    fun :(context :any, key :K, value :V, self :LRUCache<K,V>)=>void,
    context? :any,
    desc? :boolean
  ) : void;

  // Returns a JSON (array) representation
  toJSON() : Array<{key :K, value :V}>;

  // Returns a human-readable text representation
  toString() : string;
}

If you need to perform any form of finalization of items as they are evicted from the cache, wrapping the shift method is a good way to do it:

let c = new LRUCache(123);
c.shift = function() {
  let entry = LRUCache.prototype.shift.call(this);
  doSomethingWith(entry);
  return entry;
}

The internals calls shift as entries need to be evicted, so this method is guaranteed to be called for any item that's removed from the cache. The returned entry must not include any strong references to other entries. See note in the documentation of LRUCache.prototype.put (Object key, Object value) -> Object entry.

MIT license

Copyright (c) 2010-2016 Rasmus Andersson https://rsms.me/

Permission is hereby granted, free of charge, to any person obtaining a copy of this software and associated documentation files (the "Software"), to deal in the Software without restriction, including without limitation the rights to use, copy, modify, merge, publish, distribute, sublicense, and/or sell copies of the Software, and to permit persons to whom the Software is furnished to do so, subject to the following conditions:

The above copyright notice and this permission notice shall be included in all copies or substantial portions of the Software.

THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY, FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM, OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE SOFTWARE.