Skip to content

Memoization decorator for immmutable functions. Uses ES6 WeakMap to avoid memory leaks. Best for immutable arguments.

Notifications You must be signed in to change notification settings

afiedler/memoized-decorator

Folders and files

NameName
Last commit message
Last commit date

Latest commit

 

History

1 Commit
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 

Repository files navigation

memoized-decorator

This package exports a memoization function and a ECMAScript decorator that can be used to memoize methods on a class. It is primarily designed to memoize functions or methods that receive immutable data.

Here's an example of the memoization function in use:

const memoize = require('./index').memoize;

function fib(n) {
   if(n == 1 || n == 2) return 1;
   return fib(n-1) + fib(n-2);
}

const memoFib = memoize(fib);

memoFib(10);

// only calls fib(n) once for each n, since the results are memoized

Here is an example of the decorator:

const memoized = require('memoized-decorator').decorator;

class MyMath {
   @memoized
   fib(n) {
      if(n == 1 || n == 2) return 1;
      return this.fib(n-1) + this.fib(n-2);
   }
}

Unlike other memoization functions for Javascript, this function does not compute a cache key using the function arguments. Instead, it uses Map and WeakMap to form a tree which is then searched for memoized values. This should be far faster, especially if the memoized function is called with large arguments. Additionally, because WeakMap lets the garbage collector release memory if all other references to a key have been destroyed, this method should also save memory.

Caveat: Mutable Arguments

The main caveat to this method is that if you modify a function argument but do not make a deep copy, by default the memoized function will not recognize the difference. Here's an example:

const memoSum = memoize((a, b) => a.val + b.val);
const A = { val: 5 };
const B = { val: 6 };

memoSum(A,B);
// -> 11

A.val = 1;
memoSum(A,B);
// -> 11

Because of this, this memoization method is primarily for immutable data. However, if you must use your memoized function with mutable data as well, you can tell it to conditionally bypass memoization by providing an exceptWhen option.

const memoSum = memoize((a, b) => a.val + b.val, { exceptWhen: (arg) => arg.mutable });
const A = { val: 5, mutable: true };
const B = { val: 6 };

memoSum(A,B);
// -> 11  -- not memoized since exceptWhen(A) === true!

A.val = 1;
memoSum(A,B);
// -> 7

With the decorator, this can be passed in as well: @memoized({ exceptWhen: (arg) => arg.mutable }). The exceptWhen option can also take a index as the second argument, which will be set to the argument index. For example, this will not memoize when the second argument has the prop mutable: (arg, index) => index === 1 && arg.mutable.

Other Caveats

  • If the function you are memoizing returns undefined for a given set of arguments, that value will not be cached and subsequent calls to the memoized version of your function will still call the original.
  • Primitive keys (strings, numbers, booleans, symbols, and null) are stored in a Map, which means they will not be cleared from memory if they are de-referenced in outside code.
  • If your function returns large objects, those will remain in memory until the corresponding keys are de-referenced.

Other Details

The @memoized decorator and the memoize function attached the unmemoized version of the function like this:

function myFunc () {}

const memoFunc = memoize(myFunc);
memoFunc.unmemoized === func;
// -> true

The memoized function will have a name property set to the original function name plus "_memoized":

memoFunc.name;
// -> "myFunc_memoized"

About

Memoization decorator for immmutable functions. Uses ES6 WeakMap to avoid memory leaks. Best for immutable arguments.

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published