Hamt 3 - The Final Iteration

Hot on the heels of Hamt V1, here comes Hamt V2. This release focuses on superficial API compatibility with ES6’s Map, making it easier to start using immutable data structures in your code. This also means that Hamt’s finally showing Javascript iterators some love.

You can find the updated documentation on Github along with the source code. Install the latest version of Hamt by running $ npm install hamt.

Let’s take a look.


Hamt maps now implement the ES6 iterable protocol so you can use them in for of loops and with any API that takes a Javascript iterable:

const hamt = require('hamt');

const h = hamt.empty
    .set('key', 'value')
    .set('more key', 'another value');
for (let [key, value] of h)
    console.log(key, value);


Unlike hamt.fold, iterators are lazy, so you can access only the map elements you need, or break out of iteration early:

/// check that all values in `map` satisfy function `f`
const all = (pred, map) => {
    for (let [key, value] of map)
        if (!pred(value))
            return false;
    return true;

// Kind of silly to short circuit for a map of this size but,
// for large maps, this ability is important.
let h = hamt.emtpy
    .set('a', 4)
    .set('b', 84)

all(x => x < 100, h) === true;

h = h.set('c', 101);
all(x => x < 100, h) === false;

For compatibility with ES6’s Map, the keys and values functions have also been updated to return iterators instead of arrays:

const h = hamt.empty
    .set('key1', 1)
    .set('key2', 2)

// These two are the same
Array.from(h.keys()) === ['key1', 'key2']

// These two are the same
Array.from(h.values()) === [1, 2]

Keep in mind that like hamt.fold, iteration is unordered. The insertion order does not determine the order of iteration.

Other API changes


For compatibility with Map, you can now also get the number of elements in a map with the .size property.

const h = hamt.empty.set('a', 1).set('b', 2);
h.size === 2;

.size is still an alias for hamt.count(map), which is a O(N) operation (I’ll fix that one of these days…)

hamt.set argument order fixed

In Hamt V1, I reversed the argument order of hamt.set to be hamt.set(value, key, map). For some reason, yesterday me thought that it would be more common to bind the value parameter in hamt.set. That was stupid. This argument order is confusing and far less useful than I imagined. Thankfully, even V1 kept the correct order for the method based API: map.set(key, value).

It turns out that binding the key argument is actually a whole lot more helpful, so Hamt 2 restores the old hamt.set(key, value, map) argument order:

// Now, code like this is easy to write again
const setKey = hamt.set.bind(null, 'key');

myMap = setKey(4, myMap);


hamt.pairs has been replaced by hamt.entries. Unlike pairs, entries returns a Javascript iterator over the key, value pairs in a map:

const h = hamt.empty
    .set('key1', 1)
    .set('key2', 2)

const it = h.entries();
let x = it.next();
x.value === ['key1', 1]

x = it.next()
x.value === ['key2', 2]

x = it.next();
x.done === true;

hamt.forEach and map.forEach

forEach iterates over an entire map using a function:

const h = hamt.empty
    .set('key1', 1)
    .set('key2', 2)

h.forEach((value, key, map) => {
   console.log(value, key); 

In cases where you need to iterate over all values of a map, forEach is much faster than using for of or iterators (and for even better perf, use hamt.fold directly).


For a complete API reference, see the updated documentation. And if you run into any issues, feel free to file a bug or submit a pull request.

Also, despite the name of this post, this most definitely is not the final release of Hamt. After this though, we’ll have to do a reboot or something because calling the next major release Hamt 4 would just be embarrassing.