Like Underscore, but lazier

Lazy.js it a utility library for JavaScript, similar to Underscore and Lo-Dash but with one important difference: lazy evaluation (also known as deferred execution). This can translate to superior performance in many cases, especially when dealing with large arrays and/or "chaining" together multiple methods. For simple cases (map, filter, etc.) on small arrays, Lazy's performance should be similar to Underscore or Lo-Dash.

The following chart illustrates the performance of Lazy.js versus Underscore and Lo-Dash for several common operations using arrays with 10 elements each on Chrome:

You can see that the performance difference becomes much more significant for methods that don't require iterating an entire collection (e.g., indexOf, take) as the arrays get larger:

Intrigued? Great! Now let's look at how Lazy.js actually works.

Introduction

We'll start with an array containing 1000 integers. Incidentally, generating such an array using Lazy.js is quite trivial:

var array = Lazy.range(1000).toArray();

Note the toArray call; without it, what you'll get from Lazy.range won't be an actual array but rather a sequence which you can iterate over using each. But we'll get to that in a moment.

Now let's say we want to take the squares of each of these numbers, increment them, and then take the first five even results. We'll use these helper functions, to keep the code concise:

function square(x) { return x * x; }
function inc(x) { return x + 1; }
function isEven(x) { return x % 2 === 0; }

Yes, this is admittedly a very arbitrary goal. (Later I'll get around to thinking of a more realistic scenario.) Anyway, here's one way you might accomplish it using Underscore and its convenient chain method:

var result = _.chain(array).map(square).map(inc).filter(isEven).take(5).value();

This query does a lot of stuff:

• map(square): iterates over the array and creates a new 1000-element array
• map(inc): iterates over the new array, creating another new 1000-element array
• filter(isEven): iterates over that array, creating yet another new (500-element) array
• take(5): all that just for 5 elements!

So if performance and/or efficiency were a concern for you, you would probably not do things that way using Underscore. Instead, you'd likely go the procedural route:

var results = [];
for (var i = 0; i < array.length; ++i) {
  var value = (array[i] * array[i]) + 1;
  if (value % 2 === 0) {
    results.push(value);
    if (results.length === 5) {
      break;
    }
  }
}

There—now we we haven't created have any extraneous arrays, and we did all of the work in one iteration. Any problems?

Well, yeah. The main problem is that this is one-off code, which isn't reusable and took a bit of time to write. If only we could somehow leverage the expressive power of Underscore but still get the performance of the hand-written procedural solution...

---

That's where Lazy.js comes in! Here's how we'd write the above query using Lazy.js:

var result = Lazy(array).map(square).map(inc).filter(isEven).take(5);

Looks almost identical, right? That's the idea: Lazy.js aims to be completely familiar to experienced JavaScript devs. Every method from Underscore should have the same name and identical behavior in Lazy.js, except that instead of returning a fully-populated array on every call, it creates a sequence object with an each method.

What's important here is that no iteration takes place until you call each. Which means that, unlike the Underscore example above, the equivalent Lazy.js query creates no extra arrays. Essentially it combines all necessary operations into a sequence that behaves quite a bit like the procedural code we wrote a moment ago.

Of course, unlike the procedural approach, Lazy.js lets you keep your code clean and functional, and focus on buliding an application instead of optimizing array traversals.

So, cool. Is that all? I think not!

Indefinite sequence generation

The sequence-based paradigm of Lazy.js lets you do some pretty cool things that simply aren't possible with Underscore's array-based approach. One of these is the generation of indefinite sequences, which can go on forever, yet still support all of Lazy's built-in mapping and filtering capablities.

Want an example? Sure thing! Let's say we want 300 unique random numbers between 1 and 1000.

var uniqueRandsFrom1To1000 = Lazy.generate(function() { return Math.random(); })
  .map(function(e) { return Math.floor(e * 1000) + 1; })
  .uniq()
  .take(300);

// Output: see for yourself!
uniqueRandsFrom1To1000.each(function(e) { console.log(e); });

Pretty neat. How about a slightly more advanced example? Let's use Lazy.js to make a Fibonacci sequence.

var fibonacci = Lazy.generate(function() {
  var x = 1,
      y = 1;
  return function() {
    var prev = x;
    x = y;
    y += prev;
    return prev;
  };
}());

// Output: undefined
var length = fibonacci.length();

// Output: [2, 2, 3, 4, 6, 9, 14, 22, 35, 56]
var firstTenFibsPlusOne = fibonacci.map(inc).take(10).toArray();

OK, what else can we do with Lazy.js?

Asynchronous iteration

You've probably seen code snippets before that show how to iterate over an array asynchronously in JavaScript. But have you seen an example packed full of map-y, filter-y goodness like this?

// The second argument defines a 100-millisecond interval between each element.
var asyncSequence = Lazy.async(array, 100)
  .map(inc)
  .filter(isEven)
  .take(20);

// This function returns immediately and begins iterating over the sequence asynchronously.
asyncSequence.each(function(e) {
  console.log(new Date().getMilliseconds() + ": " + e);
});

All right... what else?

Event sequences

With indefinite sequences, we saw that unlike Underscore and Lo-Dash, Lazy.js doesn't actually need an in-memory collection to iterate over. And asynchronous sequences demonstrate that it also doesn't need to do all its iteration at once.

Now here's a really cool combination of these two features: with a small extension to Lazy.js (lazy.dom.js, a separate file to include in browser-based environments), you can apply all of the power of Lazy.js to handling DOM events. In other words, Lazy.js lets you think of DOM events as a sequence—just like any other—and apply the usual map, filter, etc. functions on that sequence.

Here's an example. Let's say we want to handle all mousemove events on a given DOM element, and show their coordinates in one of two other DOM elements depending on location.

// First we define our "sequence" of events.
var mouseEvents = Lazy.events(sourceElement, "mousemove");

// Map the Event objects to their coordinates, relative to the element.
var coordinates = mouseEvents.map(function(e) {
  var elementRect = sourceElement.getBoundingClientRect();
  return [
    Math.floor(e.clientX - elementRect.left),
    Math.floor(e.clientY - elementRect.top)
  ];
});

// For mouse events on one side of the element, display the coordinates in one place.
coordinates
  .filter(function(pos) { return pos[0] < sourceElement.clientWidth / 2; })
  .each(function(pos) { displayCoordinates(leftElement, pos); });

// For those on the other side, display them in a different place.
coordinates
  .filter(function(pos) { return pos[0] > sourceElement.clientWidth / 2; })
  .each(function(pos) { displayCoordinates(rightElement, pos); });

Anything else? Of course!

String processing

Now here's something you may not have even thought of: String.split. In JavaScript, this returns an array of substrings. If you think about it, this often means doing more work than necessary; but it's the quickest way (from a developer's standpoint) to get the job done.

For example, suppose you wanted the first five lines of a block of text. You could always do this with Underscore:

var firstFiveLines = _(text.split("\n")).take(5);

But of course, this actually splits the entire string into every single line.

In lower-level languages—e.g. Java, C#— we have the notion of streams. A stream is not necessarily all read into memory. We can get something like that with Lazy.js by calling Lazy.split:

var firstFiveLines = Lazy.split(text, "\n").take(5);

This way we can read the first five lines of an arbitrarily large string (without pre-populating a huge array) and map/reduce on it just as with any other sequence.

Available functions

Currently the following functions are available (meaning you can call them on any sequence, such as what you get back from Lazy(array), Lazy.generate, Lazy.range, or Lazy.async(array)).

• map
• pluck
• reduce (aka inject or foldl)
• reduceRight (aka foldr)
• filter
• reject
• where
• invoke
• find
• findWhere
• first (aka head or take)
• rest (aka tail or drop)
• initial
• last
• sortBy
• groupBy
• countBy
• uniq
• zip
• concat
• without
• difference
• union
• intersection
• flatten
• compact
• shuffle
• every (aka all)
• some (aka any)
• indexOf
• lastIndexOf
• sortedIndex
• contains
• min
• max

This library is experimental and still a work in progress.