tracked-queue

An autotracked implementation of a double-ended queue, implemented as a ring-buffer backed by a native JavaScript array.

Usage no npm install needed!

<script type="module">
  import trackedQueue from 'https://cdn.skypack.dev/tracked-queue';
</script>

README

tracked-queue

CI Supportd TypeScript Versions

An autotracked implementation of a double-ended queue, implemented as a ring-buffer backed by a native JavaScript array, with optimal performance for all common operations:

  • O(1) push and pop from either end of the queue
  • O(1) read from any element in the queue
  • O(N) iteration of the whole queue
  • O(N) access to any range of size N within the queue
  • O(N+X) storage for a queue of size N, with X a fixed overhead on the order of a few tens of bytes:
    • backing storage of size N
    • a single capacity value
    • two pointers into the storage, with the additional cost of one Glimmer "tag" for each of them

This is handy for cases where you need to be able to push items onto and pop items off of either end of a queue of a fixed size, especially where you want to be able to have fast reads from anywhere in the queue.

For example, if you have a stream of events coming across a websocket connection, which you want to display to a user, but you only ever want to keep 1,000 of them in memory at any given time, this allows you to simply push onto the back of the queue as new items come in, with no need to manually manage the front of the queue to maintain the queue size.

Contents

Example

Create a queue of a specified capacity, and then push items into it from existing arrays or individual values:

// Create a queue with capacity 5, from an existing array of elements:
import TrackedQueue from 'tracked-queue';
let queue = new TrackedQueue<string>({ capacity: 5 });

queue.append(['alpha', 'bravo', 'charlie']);
console.log(queue.size); // 3
console.log([...queue]); // ["alpha", "bravo", "charlie"]

queue.pushBack('delta');
console.log(queue.size); // 4
console.log([...queue]); // ["alpha", "bravo", "charlie", "delta"]

If you append more elements to the back of the queue, exceeding its capacity, it will drop the front of the queue, with the size of the queue never exceeding its specified capacity, and any items removed to make room are returned. (The same goes for prepend, on the front of the queue.)

let pushedOut = queue.append(["echo", "foxtrot");
console.log(queue.size); // 5
console.log([...queue]); // ["bravo", "charlie", "delta", "echo", "foxtrot"]
console.log(pushedOut); // ["alpha"]

You can also add and remove items to and from either end of the queue:

let poppedFromBack = queue.popBack();
console.log(poppedFromBack); // "foxtrot"
console.log([...queue]); // ["bravo", "charlie", "delta", "echo"]

let poppedFromFront = queue.popFront();
console.log(poppedFromFront); // "bravo"
console.log([...queue]); // ["charlie", "delta", "echo"]

queue.pushBack('golf');
queue.pushFront('hotel');
console.log([...queue]); // ["hotel", "charlie", "delta", "echo", "golf"]

These also return a value which had to be popped to make room for them, if any:

let poppedByPush = queue.pushBack('india');
console.log([...queue]); // ["charlie", "delta", "echo", "golf", "india"]
console.log(poppedByPush); // "hotel"
// make the queue non-empty
queue.popFront();
let nothingPopped = queue.pushFront('juliet');
console.log([...queue]); // ["juliet", "delta", "echo", "golf", "india"]
console.log(poppedByPush); // undefined

Compatibility

  • Ember.js v3.16 or above
  • Ember CLI v2.13 or above
  • Node.js v12 or above

TypeScript

This project follows the current draft of the Semantic Versioning for TypeScript Types proposal.

  • Currently supported TypeScript versions: v4.1, v4.2, and v4.3
  • Compiler support policy: simple majors
  • Public API: all published types not in a -private module are public

Browser support

This project uses native Proxy (via a dependency), and so is not compatible with IE11. It supports N-1 for all other browsers.

Installation

  • With npm:

    npm install tracked-queue
    
  • With yarn:

    yarn add tracked-queue
    
  • With ember-cli:

    ember install tracked-queue
    

Docs

Performance

The "dummy app" includes two performance demos, which you can see by running ember serve and navigating to http://localhost:4200:

  • A render performance demo showing that the queue implementation itself is never the bottleneck for performance; the cost of rendering DOM elements is.

  • An ops performance demo, which allows you to see the behavior of pushing and popping to and from the front and back of the queue. (This is a naive measurement using the Performance API.) A few things to observe:

    • When the queue capacity is much larger than the number of items pushed into or popped out of it, the performance of pushBack and popBack from the back of the queue is comparable to native array push and pop actions, because it is just those operations plus bumping the index for the "back" of the queue

    • When the number of items pushed exceeds the queue capacity, causing the queue to “wrap”, pushBack and popBack insertion times drops to around 100× worse than native, but space usage is bounded: native push and pop are fast because they allow the buffer to grow in an unbounded fashion. This is the fundamental tradeoff of a ring-buffer: it provides control over space usage in exchange for slightly higher costs for push and pop on the back of the queue

    • The performance of pushFront and popFront is about 10× worse than native unshift and shift up to a threshold (which varies per-browser), at which point the performance of the native methods becomes hundreds of times worse and degrades rapidly, to the point where pushFront and popFront will hang a browser tab when dealing with much more than around 100,000 operations, as the array resizing and moving around items grows in an unbounded fashion. Meanwhile, pushFront and popFront have consistent performance characteristics no matter how large the queue is.

You can see these dynamics clearly by varying the number of operations to perform and the size of the queue.

Contributing

See the Contributing guide for details.

License

This project is licensed under the BSD 2-Clause License.