js-coroutines
Supports all browsers and React Native
When is the right time to sort a massive array on the main thread of a Javascript app? Well any time you like if you don't mind the user seeing all of your animations and effects jank to hell. Even transferring to a worker thread is going to hit the main thread for serialization and stutter everything.
So when is the right time? Well it's in all those gaps where you animation isn't doing anything and the system is idle. If only you could write something to use up that time and then relinquish control to the system so it can animate and do the rest of the work, then resume in the next gap. Well now you can...
Get 60fps while sorting an array of 10 million items with js-coroutines
Quick Start
The project's main web site contains examples of js-coroutines in operation, explains how it can provide benefits to your project and has links to the full API docs plus some examples.
JS-COROUTINES Overview and API docs
How it works?
This dev.to article goes into detail about how js-coroutines works
Demo
See the Code Sandbox Demo.
Animating Using Coroutines
Another super useful way of using coroutines is to animate and control complex states - js-coroutines provides this too with the powerful
update
method that runs every frame in high priority.
There's an example of how to write your own animation later and you can see this CodeSandbox demo of stateful animations, or this game built using js-coroutines, for more.
Commonly required asynchronous operations
You can use *stringify()
and *parse()
to manipulate JSON in an idle coroutine that won't
block the main thread.
You can use stringifyAsync()
and parseAsync()
to perform JSON parsing and stringifying
anywhere you can take a promise or await
a response.
You can use *compress()
and *decompress()
to compress to storable/transmittable strings.
You can use compressAsync()
and decompressAsync()
to perform compression and decompression
anywhere you can take a promise or await
a response.
Compression
js-coroutines uses lz-string for compression.
LZ-String GitHub/Documentation.
Installation
npm install --save js-coroutines
Usage
You can make your own generator functions that do anything you like and yield
to check
if there is time remaining this frame:
import {run, sort, stringify} from 'js-coroutines'
...
let json = await run(function*() {
const results = [];
for(let i = 0; i < 10000000; i++) {
results.push(Math.random() * 10000);
//Check how much time left every 100 entries
if(i % 100 === 0) yield;
}
//Pass to a coroutine sort function
yield* sort(results, value=>value)
return yield* stringify(results);
})
Or you can just use the Async helper functions in an async routine. This is less powerful, but you don't have to start writing generator functions or working out where to yield.
import { parseAsync, mapAsync } from "js-coroutines";
async function process(url) {
const response = await fetch("someurl");
//Use the coroutine version of parse, rather than blocking
//the main thread permanently by using .json()
const result = await parseAsync(await response.text());
//Imagining the result is some database rows, map out the
//desired response without blocking the main thread for paints
const values = await mapAsync(result, (row) => ({
item: row.time,
value: row.quantity * row.unitPrice,
}));
return values;
}
Getting Started With Async Functions
Async functions are the easiest way to use js-coroutines if you just need to handle common functions like sorts, finds, filters and JSON parsing in the background. If you need to break up your own logic you will have to write a generator.
Just import the xxxAsync
version of the function from js-coroutines and
use a standard Promise chain or await
and the code will run only in the
gaps.
async function asyncFunctions() {
// Parse the JSON async
let o = await parseAsync(json);
// Concatenate arrays in the background
for (let i = 1; i < 12; i++) {
o = await concatAsync(o, o);
}
// Write out the arrays
let output = await stringifyAsync(o);
// Map ids from the array in the background
let justIds = await mapAsync(o, (v) => v.id);
// Return the JSON of just the ids
return [output, await stringifyAsync(justIds)];
}
Getting Started With Function Pipelines
You can also create pure functional pipelines using pipe, tap, branch, repeat and call
import {pipe, parseAsync, tap, mapAsync} from 'js-coroutines'
const process = pipe(
parseAsync,
function * (data) {
let i = 0
let output = []
for(let item of data) {
output.push({...item, total: item.units * item.price})
if((i++ % 100)==0) yield
}
return output
},
mapAsync.with(v=>({value: v.total, item: v.item})),
tap(console.log),
stringifyAsync
)
...
console.log(await process(data))
Getting Started Writing Your Own Generators
js-coroutines
uses generator functions and requestIdleCallback
to let you easily split up
your work with minimal effort.
A simple generator:
await run(function* () {
const strings = [];
let results;
//Create 2 million rows of random values
results = new Array(2000000);
for (let i = 0; i < 2000000; i++) {
//Every 128th record, check to see if we still have time
//run the remainder on another tick if we don't
if ((i & 127) === 0) yield;
results[i] = (Math.random() * 10000) | 0;
}
//Double all the values
yield* forEach(
results,
yielding((r, i) => (results[i] = r * 2))
);
//Get the square roots
const sqrRoot = yield* map(
results,
yielding((r) => Math.sqrt(r))
);
//Sum all of the items
const sum = yield* reduce(
results,
yielding((c, a) => c + a, 64),
0
);
//Join the arrays
yield* append(results, sqrRoot);
// Sort the results
yield* sort(results, (a, b) => a - b);
return results;
});
As you can probably see, it comes ready with the most useful functions for arrays:
forEach
map
filter
reduce
findIndex
find
some
every
sort
append
(array into array)concat
(two arrays into a new array)
The helper yielding
wraps a normal function as a generator and checks remaining time
every few iterations. You can see it in use above. It's just a helper though - if
your map
function needs to do more work it can just be a generator itself,
yield when it likes and also pass on to deeper functions that can yield:
const results =
yield *
map(inputArray, function* (element, index) {
//Every 200 indices give up work
//on this frame by yielding 'true'
//yield without true, checks the amount
//of remaining time
if (index % 200 === 199) yield true;
//Yield out a filter operation
let matched = yield* filter(
element,
yielding((c) => c > 1000)
);
//Now yield out the calculation of a sum
return yield* reduce(
matched,
yielding((c, a) => c + a),
0
);
});
yielding(fn, [optional yieldFrequency]) -> function *
Async
Former runAsync
is deprecated. You may yield a Promise instead.
js-coroutines will automatically restart the coroutine when the
Promise is resolved.
const results = await run( function* () {
const response = yield fetch("http://someurl");
const rows = yield response.json();
yield* sort(rows, (a) => a.value);
return processed;
});
Update coroutines
A great way to do stateful animation is using a coroutine running every frame.
In this case when you yield
you get called back on the next frame making
stateful animations a piece of cake:
import { update } from "js-coroutines";
//Animate using a coroutine for state
update(function* () {
while (true) {
//Move left to right
for (let x = -200; x < 200; x++) {
logoRef.current.style.marginLeft = `${x * multiplier}px`;
yield;
//Now we are on the next frame
}
//Move top to bottom
for (let y = 0; y < 200; y++) {
logoRef.current.style.marginTop = `${y * multiplier}px`;
yield;
}
//Move diagonally back
for (let x = 200; x > -200; x--) {
logoRef.current.style.marginLeft = `${x * multiplier}px`;
logoRef.current.style.marginTop = ((x + 200) * multiplier) / 2 + "px";
yield;
}
}
});
Writing Coroutines with the API
run(coroutineFunction, msToLeaveSpare=1, timeout=160) -> TerminatablePromise(Any)
coroutineFunction
must be a function *
Run your coroutine, which will occupy up to the last amount of
m/s specified in the msToLeaveSpare
(0.5 is the minimum) of the idle
time on the thread. timeout
specifies the time before it will run
if there is no idle time (default 1/10 frames).
The promise returned has a terminate(result)
function that can be used
to stop the calculation early - maybe you want to go again with different
parameters.
yield
inside your coroutine will check how much time is left and continue
if there is enough.
yield 2
yielding a number results in a check for at least that number of ms remaining.
yield true
will definitely abandon the current frames work. Useful if you
are about to/just have allocated tons of memory to give time for GC.
yield* generatorFn([param], [...param])
call a generator function which
will take over yielding time checks and return the value it creates when done.
function* myCoroutine() {
const results = [];
for (let i = 1; i < 1000000; i++) {
if ((i & 127) === 0) yield; //time check
results.push(i);
}
yield true; // end current frame processing
let anotherArray = new Array(results.length);
yield true; // give time for GC
// Run a for loop on the results
yield* forEach(
results,
yielding(
(result, index, collection) =>
(anotherArray[index] = result / collection.length)
)
);
return anotherArray;
}
*yielding(fn, [optional frequency=8]) -> function *
Converts a normal function into one that yields every frequency
calls.
Very useful for providing map/filter functions etc.
wrapAsPromise(coroutine) -> function([params]) -> Promise(Any)
Returns an async function that can be called with await and will call the passed in coroutine forwarding parameters.
//Create an async function
const toTuplesAsync = wrapAsPromise(function* (array) {
let output = [];
//Create tuples
for (let i = 0; i < array.length; i += 2) {
output.push([array[i], array[i + 1]]);
yield;
}
return output;
});
...
async function myProcess() {
const data = await getDataFromSomewhere();
//Call your wrapped coroutine
const tuples = await toTuplesAsync(data);
//do something with the result
return processTuplesSomehow(tuples);
}
License
js-coroutines - MIT (c) 2020 Mike Talbot
Timsort - MIT (c) 2015 Marco Ziccardi (c) 2020 Mike Talbot (Generator modifications)
JSON stringify - Public Domain (c) 2017 Douglas Crockford (c) 2020 Mike Talbot (Generator modifications)
JSON Parse - yastjson - MIT (c) 2020 5u9ar (zhuyingda) (c) 2020 Mike Talbot (Optimisations and generator modifications)