Memoize Immutable

An efficient memoizer for functions that only receive immutable arguments. Ideal for Redux and similar environments.

This lib is only compatible with browsers that implement WeakMap and Map natively. (which have good browser support).

How is it different from other memoizers?

In order to index cached results, most memoizers serialize arguments using JSON.stringify or similar methods. When working with immutable data, using a WeakMap based cache is much more CPU and memory efficient. This memoizer is designed to work with such caches.

Install

npm install --save memoize-immutable

API

memoize( fn [, options ] )

• fn: the function to memoize
• options (optionnal):
  • cache: a cache instance implementing .has, .get and .set methods (defaults to TupleMap)
  • limit: limit the size of the default cache (incompatible with cache option)

returns a memoized function. Note: the .displayName of the returned function will be '<original name>Memoized'.

Usage

var memoize = require('memoize-immutable');

var nbExecs = 0;
var arraySum = function(arr) {
  nbExecs++;
  return arr.reduce(function(acc, curr) {
    return acc + curr;
  }, 0);
};
var arraySumMemoized = memoize(arraySum);


var arr1 = [ 1, 2, 3, 4, 5, 6 ];
var copy = arr1;

expect(arraySumMemoized(arr1)).to.equal(21);
expect(nbExecs).to.equal(1);

expect(arraySumMemoized(copy)).to.equal(21);
expect(nbExecs).to.equal(1);

// Of course, you shouldn't mutate the arguments, or else...
arr1.push(7);
expect(arraySumMemoized(arr1)).to.equal(21);
expect(nbExecs).to.equal(1);

var clone = arr1.concat();
expect(arraySumMemoized(clone)).to.equal(28);
expect(nbExecs).to.equal(2);

## Choosing a cache store

NB: When in doubt, don't use an optional cache.

The following instructions will help choose optimal cache store for a given function. Before you proceed, make sure you know the definition of the following terms:

• primitive: Any number, string, boolean, undefined or null value is considered primitive.
• non-primitive: An object, array or function value is non-primitive.
• named arguments: here is a function that doesn't accept named arguments: drawRect(20, 50, 100, 150, '#000'); and the same function, accepting named arguments: drawRect({x: 20, y: 50, width: 100, height: 150, color: '#000'}); which is expected to have the exact same result as: drawRect({color: '#000', width: 100, height: 150, x: 20, y: 50});

1. The function accepts a single argument (not named argument, see below)
2. The function accepts a single non-primitive argument. → use a native WeakMap.
3. The function accepts a single primitive argument. → use the LRUMap (or a native Map if its size isn't a problem).
4. The function accepts multiple arguments, but the number of arguments never changes
5. The function accepts primitive arguments, always mixed with at least one non-primitive argument → use the MixedTupleMap.
6. The function only accepts non-primitive arguments. → use the WeakTupleMap.
7. The function accepts a single object of named arguments → use the NamedTupleMap.
8. In any other case → use the default TupleMap.

license

MPL-2.0

Author

@louis_remi