Background: I'm doing some user interface tests that need to detect if people are paying attention or not. But, this question is not about the page visibility API.
Specifically, I would like to know how my Javascript code will be affected if the current tab is not active, or the browser window is not active, in different browsers. I've dug up the following so far:
ios 5 pauses javascript when tab is not active
setInterval and setTimeout delay is reduced when tabs are not active - seems like this just started appearing recently and can mess up Jasmine unit tests, around other things.
requestAnimationFrame is slowed when tab is not active (reasonable, can't think of why this would affect anyone too much)
I have the following questions:
Other than mobile browsers, do desktop browsers ever pause JS execution when a tab is not active? When and which browsers?
Which browsers reduce the setInterval repeat? Is it just reduced to a limit or by a percentage? For example, if I have a 10ms repeat versus a 5000ms repeat, how will each be affected?
Do these changes happen if the window is out of focus, as opposed to just the tab? (I imagine it would be harder to detect, as it requires the OS API.)
Are there any other effects that would not be observed in an active tab? Could they mess things up that would otherwise execute correctly (i.e. the aforementioned Jasmine tests)?
setInterval
/setTimeout
times under 1000ms are changed to 1000ms when the tab/window is blurred
setInterval
/setTimeout
times under 1000ms are changed to 1000ms when the tab/window is blurred. Not clear what you tried to convey
Test One
I have written a test specifically for this purpose:
Frame Rate Distribution: setInterval vs requestAnimationFrame
Note: This test is quite CPU intensive. requestAnimationFrame
is not supported by IE 9- and Opera 12-.
The test logs the actual time it takes for a setInterval
and requestAnimationFrame
to run in different browsers, and gives you the results in the form of a distribution. You can change the number of milliseconds for setInterval
to see how it runs under different settings. setTimeout
works similarly to a setInterval
with respect to delays. requestAnimationFrame
generally defaults to the 60fps depending on the browser. To see what happens when you switch to a different tab or have an inactive window, simply open the page, switch to a different tab and wait for a while. It will continue to log the actual time it takes for these functions in an inactive tab.
Test Two
Another way to test it is to log the timestamp repeatedly with setInterval
and requestAnimationFrame
and view it in a detached console. You can see how frequently it is updated (or if it is ever updated) when you make the tab or window inactive.
Test for setInterval
Test for requestAnimationFrame
Results
Chrome
Chrome limits the minimum interval of setInterval
to around 1000ms when the tab is inactive. If the interval is higher than 1000ms, it will run at the specified interval. It does not matter if the window is out of focus, the interval is limited only when you switch to a different tab. requestAnimationFrame
is paused when the tab is inactive.
// Provides control over the minimum timer interval for background tabs.
const double kBackgroundTabTimerInterval = 1.0;
https://codereview.chromium.org/6546021/patch/1001/2001
Firefox
Similar to Chrome, Firefox limits the minimum interval of setInterval
to around 1000ms when the tab (not the window) is inactive. However, requestAnimationFrame
runs exponentially slower when the tab is inactive, with each frame taking 1s, 2s, 4s, 8s and so on.
// The default shortest interval/timeout we permit
#define DEFAULT_MIN_TIMEOUT_VALUE 4 // 4ms
#define DEFAULT_MIN_BACKGROUND_TIMEOUT_VALUE 1000 // 1000ms
https://hg.mozilla.org/releases/mozilla-release/file/0bf1cadfb004/dom/base/nsGlobalWindow.cpp#l296
Internet Explorer
IE does not limit the delay in setInterval
when the tab is inactive, but it pauses requestAnimationFrame
in inactive tabs. It does not matter whether the window is out of focus or not.
Edge
Starting from Edge 14, setInterval
is capped at 1000ms in inactive tabs. requestAnimationFrame
is always paused in inactive tabs.
Safari
Just like Chrome, Safari caps setInterval
at 1000ms when the tab is inactive. requestAnimationFrame
is paused as well.
Opera
Since the adoption of the Webkit engine, Opera exhibits the same behavior as Chrome. setInterval
is capped at 1000ms and requestAnimationFrame
is paused when the tab is inactive.
Summary
Repeating intervals for inactive tabs:
setInterval requestAnimationFrame Chrome 9- not affected not supported 10 not affected paused 11+ >=1000ms paused Firefox 3- not affected not supported 4 not affected 1s 5+ >=1000ms 2ns (n = number of frames since inactivity) IE 9- not affected not supported 10+ not affected paused Edge 13- not affected paused 14+ >=1000ms paused Safari 5- not affected not supported 6 not affected paused 7+ >=1000ms paused Opera 12- not affected not supported 15+ >=1000ms paused
What I observed : on inactive tabs in Chrome, all your setTimeout
(must be the same for setInterval
) waiting less than 1000ms are rounded to 1000ms. I think longer timeouts are not modified.
Seems to be the behavior since Chrome 11 and Firefox 5.0 : https://developer.mozilla.org/en-US/docs/DOM/window.setTimeout#Inactive_tabs
Furthermore, I don't think it behaves this way when the whole window is inactive (but it seems quite easy to investigate).
focus
and blur
events seem to detect both tab and window switches, so it could conceivably work both ways. But I wonder how the window detects if it's actually visible or not.
A newer answer to complement these: on chrome 78.0.3904.108 I notice all of these timeouts (not just those below 1000ms) taking a bit longer than expected when I move to a different tab, and then come back. The behaviour I am seeing is more correctly described as "All timeouts on inactive tabs may be delayed by some additional amount, to a maximum of 1000ms." Try running the following and switching to another tab!
let timeouts = [ 500, 1000, 2000, 3000, 10000 ]; let minExcess = document.getElementsByClassName('minExcess')[0]; timeouts.forEach(ms => { let elem = document.getElementsByClassName(`t${ms}`)[0]; let cnt = 0; let lastMs = +new Date(); let f = () => { let curMs = +new Date(); let disp = document.createElement('p'); let net = curMs - lastMs; lastMs = curMs; setTimeout(f, ms); if (minExcess.value && (net - ms) < parseInt(minExcess.value)) return; disp.innerText = `${net},`; elem.appendChild(disp); if (++cnt > 10) elem.firstElementChild.remove(); }; setTimeout(f, ms); }); body { font-size: 80%; } div { max-height: 80px; overflow-x: auto; background-color: rgba(0, 0, 0, 0.1); margin-bottom: 2px; white-space: nowrap; } p { margin: 0; } div > p { margin: 0; display: inline-block; vertical-align: top; margin-right: 2px; } input { margin: 0 0 10px 0; } .t500:before { display: block; content: '500ms'; font-weight: bold; } .t1000:before { display: block; content: '1000ms'; font-weight: bold; } .t2000:before { display: block; content: '2000ms'; font-weight: bold; } .t3000:before { display: block; content: '3000ms'; font-weight: bold; } .t10000:before { display: block; content: '10000ms'; font-weight: bold; }
Ignore any values delayed by less than this amount:
Success story sharing
setInterval
andrequestAnimationFrame
?setInterval
andrequestAnimationFrame
. The thing I know about is thatsetTimeout
behaves similarly tosetInterval
, in that they both have the same minimum background interval in Firefox and Chrome, and no apparent limit in other browsers.about:config
in the browser and changing thedom.min_background_timeout_value
value to something else than 1000.requestAnimationFrame
is called if the user simply switches application (Alt+Tab out of Chrome). As long as the tab is active in Chrome the "frame rate" is more or less constant.