package lru

  1. Overview
  2. Docs

Scalable LRU caches

Lru provides size-bounded finite maps that remove the least-recently-used (LRU) bindings in order to maintain the size constraint. Two implementations are provided: one is functional, the other imperative.

The functional map is backed by a priority search queue. Operations on individual elements are O(log n).

The mutable map is backed by the standard Hashtbl paired with a doubly-linked list. Operations on individual elements incur an O(1) overhead on top of hash table access.

Both versions support differentially weighted bindings, and have a capacity parameter that limits the combined weight of the bindings. To limit the maps by the number of bindings, use let weight _ = 1.

v0.2.0 — homepage

Lru

module type Weighted = sig ... end

Signature of types with measurable weight.

module F : sig ... end

Functional LRU map.

module M : sig ... end

Mutable LRU map.

One-off memoization

val memo : ?hashed:(('a -> int) * ('a -> 'a -> bool)) -> ?weight:('b -> int) -> cap:int -> (('a -> 'b) -> 'a -> 'b) -> 'a -> 'b

memo ?hashed ?weight ~cap f is a new memoized instance of f, using LRU caching. f is an open recursive function of one parameter.

~hashed are hashing and equality over the arguments 'a. It defaults to (Hashtbl.hash, Pervasives.(=)).

~weight is the weighting function over the results 'b. It defaults to fun _ -> 1.

~cap is the total cache capacity.

  • raises Invalid_argument

    when cap < 0.