w3c / IntersectionObserver

Intersection Observer

Home Page:https://www.w3.org/TR/intersection-observer/

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Polyfill: callback is called synchronously

Jessidhia opened this issue · comments

The intersection observer's callback should be called asynchronously according to spec.

https://wicg.github.io/IntersectionObserver/#update-intersection-observations-algo step 2.12

https://wicg.github.io/IntersectionObserver/#queue-an-intersectionobserverentry step 3

https://wicg.github.io/IntersectionObserver/#queue-intersection-observer-task step 3

The this._callback invocation should be done in a requestIdleCallback or setTimeout call.

I am getting an issue where both in IE and FireFox, when the polyfill is added, it quickly freezes up the page and the browser goes into "not responding". Would this synchronous callback create this behavior?

I haven't been able to test extensively so I didn't experience a freeze up of the page in my case with several IO targets being observed. Whenever I am using the polyfill I make sure I am wrapping my calls with requestIdleCallback, which indeed seems to be the best way to prevent potential "jank".

So to wrap your individual invocations, here's a snippet you can use for the time-being:

import { requestIdleCallback } from 'request-idle-callback'; // optional polyfill

const hasNativeSupport = 'IntersectionObserver' in window && window.IntersectionObserver.prototype.toString().contains('[object IntersectionObserver');
const ensureAsync = (callback) => hasNativeSupport ? callback : (...args) => requestIdleCallback(() => { callback(...args); });

export default ensureAsync;

And here's a couple of Profile snapshots taken with Firefox's devtools - bare in mind it's me scrolling in the same page up and down without a given order, so it's far from accurate to measure - I did however obtain the best result 2-3 times in a row with requestIdleCallback:

Without RIC

without ric

min 10fps

With IRC

with ric

min 20fps

Given Firefox's native implementation, I'm going to optimistically close this issue. mlisook/plastic-image#19 (comment) is also positive to see. If someone is still experiencing an issue, please comment and we can re-open.