1317422 - (globalThis) Implement "globalThis" proposal

Status: RESOLVED FIXED

(Core :: JavaScript: Standard Library, enhancement, P5)

Description

[André Bargull [:anba]]

"global" [1] is currently a stage 3 proposal [2, 3].

test262 already tests for the new "global" property [4] and we're currently failing some tests because of this.


[1] https://github.com/tc39/proposal-global
[2] https://github.com/tc39/proposals
[3] https://github.com/tc39/ecma262/pull/702
[4] https://github.com/tc39/test262/pull/765

Comment 1

[Jeff Walden [:Waldo]]

Attachment: Patch

#jslang suckered me into it.

Comment 2

[ljharb]

<3 Thanks!

Comment 3

[André Bargull [:anba]]

Comment on attachment 8815143 [details] [diff] [review]
Patch

Review of attachment 8815143 [details] [diff] [review]:
-----------------------------------------------------------------

Let's find out if this breaks any websites! :-D

::: js/src/tests/ecma_2017/Global/global-property-enumeration.js
@@ +28,5 @@
> +assertEq(desc.configurable, true);
> +assertEq(desc.writable, true);
> +assertEq(desc.value, this);
> +
> +global = 42;

The other test case already tests that `global = 42` works, should this one instead test with `this.global = 42`?

Comment 4

[Jeff Walden [:Waldo]]

Patch works fine for shell, doesn't work for browser.  Turns out I need a ToWindowProxyIfWindow in there.  But even that's not enough, because in the browser embedding, nsGlobalWindow::SetNewDocument invokes Object class init (in the process of setting up the window's prototype chain) before it creates the WindowProxy wrapper and calls js::SetWindowProxy to store the wrapper in the WINDOW_PROXY slot in the Window.

The right fix to me seems to be to refactor things so the slot's filled early enough that Object class init doesn't have these oddities to deal with.  But SetNewDocument looks an awful lot like Mos Eisley -- a wretched hive of scum and villainy -- and doesn't look particularly refactorable, nor does it seem like a good use of time to do so.  (I want to refactor global object creation at *some* point, so that embeddings can basically declaratively specify what they want their global object to look like, including its prototype chain.  So I'll probably get back to this eventually.  Just, not now.)

With the Object init approach out, I stole the approach evilpie mentioned on IRC (aside: it's truly awful that there are multiple ways to define a new one-off global property), fixed the one moderately obvious bug in it, and am tryservering it now.  Will post for a fresh review if it comes back okay.

https://treeherder.mozilla.org/#/jobs?repo=try&revision=a7cf02615566f272bd0ad7a89c865262a2a2ce62

Comment 5

[Jeff Walden [:Waldo]]

Attachment: Patch, v2

Pretty sure jandem dealt with the ToWindow* gunk at one point, so let's throw this at him this time around.

Comment 6

[Jan de Mooij [:jandem]]

Comment on attachment 8816662 [details] [diff] [review]
Patch, v2

Review of attachment 8816662 [details] [diff] [review]:
-----------------------------------------------------------------

LGTM.

Comment 7

[Tom S]

Reminder to land this.

Comment 8

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/d37fc752b0af
Implement a global 'global' property whose value is the global object.  r=jandem

Comment 9

[Wes Kocher (:KWierso) (Not reading bugmail; email directly if needed)]

https://hg.mozilla.org/mozilla-central/rev/d37fc752b0af

Comment 10

[ljharb]

Yay! When will this land in nightly and stable?

Comment 11

[Jan de Mooij [:jandem]]

(In reply to ljharb from comment #10)
> Yay! When will this land in nightly and stable?

It will be in today's Nightly. Firefox 53 will be released April 18th, according to https://wiki.mozilla.org/RapidRelease/Calendar

Comment 12

[Tom S]

Let's back this out. This is not web-compatible.

Comment 13

[Tom S]

Attachment: Backout

Comment 14

[Jeff Walden [:Waldo]]

Comment on attachment 8823427 [details] [diff] [review]
Backout

Review of attachment 8823427 [details] [diff] [review]:
-----------------------------------------------------------------

::: js/src/jsapi.cpp
@@ +1076,5 @@
>      // Object.prototype yet. Force it now.
> +    if (!global->getOrCreateObjectPrototype(cx))
> +        return false;
> +
> +    return true;

Mind also landing a separate patch, rs=me, that undoes this undoing, and keeps |global| as a Handle?

Comment 15

[ljharb]

Is there a way we could still include `global` behind a flag in nightly, so that continued testing can be done as sites fix their issues?

Comment 16

[Jeff Walden [:Waldo]]

Sure, we could add something to CompartmentOptions to frob it on, easily enough.  Invoking the frob other than as an all-or-nothing approach, tho, seems difficult.  I'm not sure exactly how useful such a thing is.  We've used such frobs to feature-guard stuff in progress, like some Intl functionality, but that was only with the aim of having an opt-in in shell tests to enable it.  And obviously, we can't expose such a thing to web content.

Comment 17

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/aa5f92e32f1d
Backout the global property propsal. r=Waldo
https://hg.mozilla.org/integration/mozilla-inbound/rev/89388a238c25
Keep nicer code in JS_ResolveStandardClass. rs=Waldo

Comment 18

[Alexandre Folle de Menezes]

How about enabling this just for 'strict' contexts, wouldn't it be more compatible ir current web content?

Comment 19

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/aa5f92e32f1d
https://hg.mozilla.org/mozilla-central/rev/89388a238c25

Comment 20

[Karl Dubost💡 :karlcow]

After implementing it in Safari Tech Preview 21
> Added support for the global property on the global object (r210052)
https://webkit.org/blog/7265/release-notes-for-safari-technology-preview-21/

It has been reverted for Web Compatibility reasons
See https://bugs.webkit.org/show_bug.cgi?id=166915

Ryosuke Niwa said
> We had to rollout this feature it broke Polymer tests.
https://bugs.webkit.org/show_bug.cgi?id=165171#c22

Comment 21

[Tom S]

From Alexandre Folle de Menezes in bug 1496748
> The 'global' proposal was renamed to 'globalThis'.  I believe this was
> already implemented with the old name, but disabled due to compatibility
> issues.
> 
> Chrome 70 already has the renamed implementation.

Comment 23

[Tom S]

Should be easy to implement this. I would volunteer to do it.

Comment 24

[Tom S]

Oh the original patch was actually not quite correct: We would resolve "global" again after it was deleted. Thanks test262 for catching this!

Comment 25

[Tom S]

Attachment: Bug 1317422 - Implement JavaScript globalThis proposal. r?jandem

Comment 26

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/autoland/rev/1e6a2dbfa104
Implement JavaScript globalThis proposal. r=jandem

Comment 27

[Andreea Pavel [:apavel]]

https://hg.mozilla.org/mozilla-central/rev/1e6a2dbfa104

Comment 28

[Chris Mills [:cmills]]

Note to MDN writers:

I have added to note to the Fx65 rel notes to cover this:
https://developer.mozilla.org/en-US/docs/Mozilla/Firefox/Releases/65#JavaScript

It would be great to get this documented fairly promptly in the new year. The Trello card is here:
https://trello.com/c/aRcYBQqA/150-globalthis-fx-65

Awesomely enough, we already seem to have it in Chinese, added by a volunteer:
https://developer.mozilla.org/zh-CN/docs/Web/JavaScript/Reference/Global_Objects/globalThis

Comment 29

[Florian Scholz (Open Web Docs)]

We need to do the following tasks on MDN:

Add browser compat data:

• Add "globalThis" to https://github.com/mdn/browser-compat-data/blob/master/javascript/builtins/globals.json

Add reference docs:

• Create https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/globalThis
  • Can be modeled after e.g. https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/undefined
  • Should talk about globalthis, self, this, window.
  • Examples could be global with web workers, global in modules, ...
• Mention "globalThis" on https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects under "Value properties"

Add interactive example:

• The folder is https://github.com/mdn/interactive-examples/tree/master/live-examples/js-examples/globalprops
• Add a new globalThis example and embed it in the globalThis reference page.

Links from other docs

We should probably link to the globalThis page from these MDN pages:

• https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Operators/this
• https://developer.mozilla.org/en-US/docs/Web/API/Window
• https://developer.mozilla.org/en-US/docs/Web/API/WorkerGlobalScope/self

Comment 30

[Florian Scholz (Open Web Docs)]

The docs are put in place thanks to Helen Durrant, https://twitter.com/goodforenergy

https://developer.mozilla.org/en-US/docs/Mozilla/Firefox/Releases/65#JavaScript
https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/globalThis

Reviews, comments, and tweaks to the MDN wiki pages are welcome (as always).