Implement undefined-like exotic objects #673

bterlson · 2016-08-17T19:28:57Z

Closes #668. Still not convinced we want this in 262 but here it is so we can discuss!

UltCombo · 2016-08-17T20:00:26Z

spec.html

@@ -4074,6 +4074,8 @@
          1. Return the result of performing Strict Equality Comparison _x_ === _y_.
        1. If _x_ is *null* and _y_ is *undefined*, return *true*.
        1. If _x_ is *undefined* and _y_ is *null*, return *true*.
+        1. If _x_ is an undefined-like exotic object and _y_ *null*, *undefined*, or an undefined-like exotic object, return *true*.


and y null

Missing verb ("is").

littledan · 2016-08-17T20:04:32Z

LGTM

ljharb · 2016-08-17T20:21:31Z

spec.html

@@ -4072,8 +4072,7 @@
      <emu-alg>
        1. If Type(_x_) is the same as Type(_y_), then
          1. Return the result of performing Strict Equality Comparison _x_ === _y_.
-        1. If _x_ is *null* and _y_ is *undefined*, return *true*.
-        1. If _x_ is *undefined* and _y_ is *null*, return *true*.
+        1. If _x_ is either *null*, *undefined*, or an undefined-like exotic object and _y_ is either *null*, *undefined*, or an undefined-like exotic object, return *true*.


Would this benefit from a new abstract operation that emulates == null, and encapsulates all three of these comparisons?

The proposed semantics seems bad to me, because it means that document.all’s from distinct frames will be reputed equal. ~~Not unrelatedly, it breaks the invariant x === y iff typeof x == typeof y && x == y.~~ (nevermind) We should optimise for sanity rather than conciseness:

If x is undefined/null and y is undefined/null, return true.

If x is a ULEO and y is undefined/null, return true.

If x is undefined/null and y is a ULEO, return true.

I was thinking if IsNullish(x) and IsNullish(y), return true. and define IsNullish as "if x is null, undefined, or an ULEO, return true. else, return false".

@ljharb In case it was not clear, an abstract operation doesn’t change my “document.all’s from distinct frames appear equal” issue.

@claudepache Oh I commented on this before on IRC. Because of Step 1. different objects are always regarded as different, because they have the same type and the strict equality operation doesn't care about "undefined-like".

Because of Step 1. different objects are always regarded as different, because they have the same type and the strict equality operation doesn't care about "undefined-like".

Ah, ok. However, I'm not sure it is good to sacrifice clarity for conciseness. The formulation of the current spec text:

If x is null and y is undefined, return true.
If x is undefined and y is null, return true.

has the advantage of not interfering with step 1. So, just adding:

If x is an undefined-like exotic object and y is undefined or null, return true.
If x is undefined or null and y is an undefined-like exotic object, return true.

seems clearer to me.

Agree very much that this is superior given the confusion. Thanks!!

chicoxyzzy · 2016-08-17T21:24:25Z

~~Isn't it better to use something like (currently inactive) proposal Typed Objects (which makes operators overloading possible) for this type of things?~~

nevermind

claudepache · 2016-08-18T18:00:51Z

spec.html

+                Object (undefined-like exotic object)
+              </td>
+              <td>
+                `"undefined"`


That row must go before the “Object (implements [[Call]]): "function"” row, because document.all is callable. :-(

Good call, thanks :-P

bterlson · 2016-08-19T17:50:01Z

Fixed the issues @claudepache raised.

bterlson · 2016-09-27T21:44:13Z

Review feedback: HTML Document Dot All Exotic Object.
Put in Annex B.

mathiasbynens · 2016-09-28T00:51:27Z

spec.html

+      <h1>Undefined-like Exotic Objects</h1>
+      <p>An <dfn>undefined-like exotic object</dfn> is an exotic object that behaves like undefined in the <emu-xref href="#sec-toboolean">ToBoolean</emu-xref> and <emu-xref href="#sec-abstract-equality-comparison">Abstract Equality Comparison</emu-xref> abstract operations and when used as an operand for the <emu-xref href="#sec-typeof-operator">typeof operator</emu-xref>. Undefined-like exotic objects have all the ordinary object internal methods and slots.</p>
+      <emu-note>
+        <p>Undefined-like exotic objects are never created by this specification. However, the <a href="https://html.spec.whatwg.org/multipage/obsolete.html#dom-document-all">document.all object</a> in web browsers is a host-created undefined-like exotic object that exists for web compatibility purposes. There are no other known examples of this type of object.</p>


Nit: document.all → <code>document.all</code>

annevk · 2017-08-12T11:18:08Z

It seems the "needs consensus" label can be removed?

littledan · 2017-08-15T13:47:41Z

Well, we have consensus for something, but not for this patch. The consensus is described in this comment.

annevk · 2017-08-16T15:03:03Z

@bterlson could you maybe up the priority for this a bit? There's a plan to touch HTMLAllCollection again (to move over legacycaller from IDL to HTML) and it'd be nice to fix this at the same time.

Also, will the <dfn> that's being added get an ID or would that need to be added manually? It'd probably be useful to have something to point to directly.

bterlson · 2017-08-16T16:55:54Z

@annevk I will work on this today. Was really hoping we could get away without this but alas. It'll be kinda finicky to move to Annex B (mostly in turning algorithm changes into Annex B mods) but I should be able to complete it today for your review.

The dfn will link to the parent clause ID without its own ID. Would that work for you?

annevk · 2017-08-16T17:05:59Z

Yeah, that's probably good enough. Still need to figure out exactly how the layering will work.

bterlson · 2017-08-17T00:42:44Z

Rebased, renamed, moved to Annex B with diff pointers. @annevk care to review?

michaelficarra · 2017-08-17T03:27:05Z

spec.html

+        <h1>Changes to Abstract Equality Comparison</h1>
+        <p>The following steps are inserted after step 3 of the <emu-xref href="#sec-abstract-equality-comparison">Abstract Equality Comparison</emu-xref> algorithm:</p>
+        <emu-alg>
+          1. If _x_ is an undefined-like exotic object and _y_ is either *null* or *undefined*, return *true*.


HTML document dot all exotic object

bterlson · 2017-08-17T21:35:56Z

After talking with @domenic, I made some changes. Biggie is instead of defining a separate object, I simply say that there exists a internal slot of a particular name and the annex B algorithms check for this internal slot rather than doing a "type check". LMK your thoughts. /cc @annevk

bterlson · 2017-08-17T21:38:25Z

spec.html

+      <emu-annex id="sec-html-document-dot-all-exotic-object-aec">
+        <h1>Changes to Abstract Equality Comparison</h1>
+        <p>The following steps are inserted after step 3 of the <emu-xref href="#sec-abstract-equality-comparison">Abstract Equality Comparison</emu-xref> algorithm:</p>
+        <emu-alg start=4>


hmm I thought this did something but it doesn't. Guess I'll delete it.

domenic

Lovely, just some nits.

domenic · 2017-08-17T21:36:55Z

spec.html

@@ -39904,6 +39904,60 @@ <h1 id="ao-issuperreference">IsSuperReference ( _V_ )</h1>
        1. Return ? ForIn/OfBodyEvaluation(|BindingIdentifier|, |Statement|, _keyResult_, ~enumerate~, ~varBinding~, _labelSet_).
      </emu-alg>
    </emu-annex>
+
+    <emu-annex id="sec-ishtmldda-internal-slot">
+      <h1>The [[isHTMLDDA]] Internal Slot</h1>


Uppercase to match other internal slot names?

domenic · 2017-08-17T21:37:49Z

spec.html

+
+    <emu-annex id="sec-ishtmldda-internal-slot">
+      <h1>The [[isHTMLDDA]] Internal Slot</h1>
+      <p>An <dfn>[[isHTMLDDA]] internal slot</dfn> may exist on implementation-defined exotic objects. Objects with an [[isHTMLDDA]] intenral slot behave like undefined in the <emu-xref href="#sec-toboolean">ToBoolean</emu-xref> and <emu-xref href="#sec-abstract-equality-comparison">Abstract Equality Comparison</emu-xref> abstract operations and when used as an operand for the <emu-xref href="#sec-typeof-operator">typeof operator</emu-xref>.</p>


Typo "intenrel"

typeof still not codeified. Actually just using title="" attribute might work better? For most of these xrefs? Maybe even it's not required at all for ToBoolean?

bterlson · 2017-08-17T21:42:49Z

@domenic ok nits should be fixed

domenic · 2017-08-17T21:45:41Z

spec.html

+
+    <emu-annex id="sec-IsHTMLDDA-internal-slot">
+      <h1>The [[IsHTMLDDA]] Internal Slot</h1>
+      <p>An <dfn>[[IsHTMLDDA]] internal slot</dfn> may exist on implementation-defined exotic objects. Objects with an [[IsHTMLDDA]] internal slot behave like undefined in the <emu-xref href="#sec-toboolean">ToBoolean</emu-xref> and <emu-xref href="#sec-abstract-equality-comparison">Abstract Equality Comparison</emu-xref> abstract operations and when used as an operand for the <emu-xref href="#sec-typeof-operator">`typeof` operator</emu-xref>.</p>


*undefined* I think?

bterlson · 2017-08-17T21:49:26Z

@annevk seem ok to you? Seems like I can't re-request a review.

claudepache · 2017-08-18T06:16:45Z

spec.html

+
+    <emu-annex id="sec-IsHTMLDDA-internal-slot">
+      <h1>The [[IsHTMLDDA]] Internal Slot</h1>
+      <p>An <dfn>[[IsHTMLDDA]] internal slot</dfn> may exist on implementation-defined exotic objects. Objects with an [[IsHTMLDDA]] internal slot behave like *undefined* in the <emu-xref href="#sec-toboolean">ToBoolean</emu-xref> and <emu-xref href="#sec-abstract-equality-comparison">Abstract Equality Comparison</emu-xref> abstract operations and when used as an operand for the <emu-xref href="#sec-typeof-operator">`typeof` operator</emu-xref>.</p>


It should be “object(s)” rather than “exotic object(s)” (here and everywhere below).

Although the behaviour described in this section is very “exotic” according the common meaning of that word, it does not make an exotic object in the sense of the definition given by the spec.

I think it's still fair, since HTMLAllCollection is an exotic object (due to custom [[Call]] and some other stuff) and it can only be used there.

I think @claudepache is right - the object may be exotic depending on host choices, but it's confusing to say its exotic when it doesn't define anything exotic. Might trigger someone to go sleuthing for hours trying to find the missing exoticness. I'll fix it.

annevk · 2017-08-18T06:59:17Z

spec.html

+      <h1>The [[IsHTMLDDA]] Internal Slot</h1>
+      <p>An <dfn>[[IsHTMLDDA]] internal slot</dfn> may exist on implementation-defined exotic objects. Objects with an [[IsHTMLDDA]] internal slot behave like *undefined* in the <emu-xref href="#sec-toboolean">ToBoolean</emu-xref> and <emu-xref href="#sec-abstract-equality-comparison">Abstract Equality Comparison</emu-xref> abstract operations and when used as an operand for the <emu-xref href="#sec-typeof-operator">`typeof` operator</emu-xref>.</p>
+      <emu-note>
+        <p>Objects with an [[IsHTMLDDA]] internal slot are never created by this specification. However, the <a href="https://html.spec.whatwg.org/multipage/obsolete.html#dom-document-all"><code>document.all</code> object</a> in web browsers is a host-created exotic object with this slot that exists for web compatibility purposes. There are no other known examples of this type of object and implementations should not create any with the exception of `document.all`.</p>


Wouldn't it be better to point to HTMLAllCollection here? document.all is a property.

Also, why not must not?

I think document.all is better. There's nothing special about HTMLAllCollection itself, really. Its instances are the weird ones. And a good way of collectively referring to all such instances is probably "document.all".

@annevk I only went with "should" because this is a note and it seems odd to use strong language in a non-normative section.

annevk · 2017-08-18T07:00:15Z

@bterlson thanks, it does.

domenic · 2017-08-18T21:14:18Z

Woohoo, thank you!!

Jack-Works · 2020-05-12T04:27:37Z

To let it evaluate to false in every case is harmful to web security. I have a suggestion that keeps the web compatibility meanwhile reduce the harm of falsy document.all.

tc39/proposal-preserve-virtualizability#2 (comment)

claudepache · 2020-05-12T09:18:10Z

@Jack-Works See 12.3.6.1 step 6 for how direct eval is specced. If you want to do the same thing for document.all, you need to add similar logic in the runtime semantics of:

the typeof operator;
operators based on boolean logic: ||, &&, ? :, !;
operator allowing comparison with null: ==;
statements making use of boolean logic: if, while, do/while, for.

(Fortunately, the newer ?. and ?? operators do not have special treatment of document.all.)

Jack-Works · 2020-05-12T09:20:12Z

Yes, but it will no longer be a threat to the security.

claudepache · 2020-05-12T12:27:34Z

In fact, I’m not exactly sure that the approach I had in mind in my previous comment would work or would be the best. But more to the point:

The current PR was specifically to sanction what browsers and the HTML spec were already implementing. That is, they felt it was better that the ES spec provided the necessary hook for the HTML spec, rather than the HTML spec monkey-patched the ES spec.

What you are proposing, is modifying the semantics of document.all in a nontrivial way, which implies implementation work and BC concern. If you feel that it is important, it should be a separate issue, and probably indeed a full-fledged proposal.

bterlson force-pushed the uleo branch 2 times, most recently from 525d10a to 2d55bef Compare August 17, 2016 19:49

UltCombo reviewed Aug 17, 2016
View reviewed changes

bterlson force-pushed the uleo branch from 2d55bef to 45d62f7 Compare August 17, 2016 20:02

bterlson mentioned this pull request Aug 17, 2016

document.all willful violation #668

Closed

ljharb reviewed Aug 17, 2016
View reviewed changes

claudepache reviewed Aug 18, 2016
View reviewed changes

bterlson added the needs consensus This needs committee consensus before it can be eligible to be merged. label Aug 18, 2016

bterlson mentioned this pull request Sep 13, 2016

Layering: add HostObjectDefinePropertyReturnFalse #688

Closed

mathiasbynens reviewed Sep 28, 2016

View reviewed changes

ljharb added has consensus This has committee consensus. and removed needs consensus This needs committee consensus before it can be eligible to be merged. labels Aug 12, 2017

annevk mentioned this pull request Aug 14, 2017

[[Construct]] for legacy platform objects? whatwg/webidl#407

Closed

littledan added the needs editorial changes label Aug 15, 2017

bterlson added 4 commits August 16, 2017 16:45

Normative: Add undefined-like exotic objects (aka document.all)

485b19e

Move ULEOs up in the typeof table

a220ecd

Refactor steps for abstract equality comparison

7dbcec4

move to annex B, rename

0867cfa

bterlson force-pushed the uleo branch from ada35ad to 0867cfa Compare August 17, 2017 00:42

michaelficarra reviewed Aug 17, 2017

View reviewed changes

bterlson commented Aug 17, 2017

View reviewed changes

domenic approved these changes Aug 17, 2017

View reviewed changes

bterlson added 3 commits August 17, 2017 14:39

Uppercase and typos

3e419ce

Codeify another instance of 'typeof'

5b82c3c

remove start

3988f9e

domenic reviewed Aug 17, 2017

View reviewed changes

domenic approved these changes Aug 17, 2017

View reviewed changes

undefined --> *undefined*

7007015

claudepache reviewed Aug 18, 2017

View reviewed changes

annevk reviewed Aug 18, 2017

View reviewed changes

bakkot added the needs test262 tests The proposal should specify how to test an implementation. Ideally via github.com/tc39/test262 label Aug 18, 2017

bakkot mentioned this pull request Aug 18, 2017

Coverage: HTML Document Dot All Exotic Object tc39/test262#1184

Closed

Replace most usages of exotic object with just object

0a1ee49

bterlson merged commit a3e06cb into tc39:master Aug 18, 2017

This was referenced Sep 7, 2017

Add a [[IsHTMLDDA]] internal slot to document.all whatwg/html#3015

Closed

buffer: improve Buffer.from performance nodejs/node#15178

Closed

bakkot removed the needs test262 tests The proposal should specify how to test an implementation. Ideally via github.com/tc39/test262 label Sep 12, 2017

claudepache mentioned this pull request Jan 18, 2020

Use a less generic name for the IsLeakableFunction abstract operation claudepache/es-legacy-function-reflection#7

Closed

ExE-Boss mentioned this pull request Feb 19, 2020

Is there also a nullish AND? tc39/proposal-nullish-coalescing#61

Closed

Jack-Works mentioned this pull request May 12, 2020

document.all is not virtualizable tc39/proposal-preserve-virtualizability#2

Open

petamoriken mentioned this pull request Oct 14, 2024

window.Promise broken: Feedback on window being removed. denoland/deno#25797

Closed

Implement undefined-like exotic objects #673

Implement undefined-like exotic objects #673

Conversation

bterlson commented Aug 17, 2016

Choose a reason for hiding this comment

littledan commented Aug 17, 2016

Choose a reason for hiding this comment

claudepache Aug 19, 2016 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

chicoxyzzy commented Aug 17, 2016 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

bterlson commented Aug 19, 2016

bterlson commented Sep 27, 2016

Choose a reason for hiding this comment

annevk commented Aug 12, 2017

littledan commented Aug 15, 2017

annevk commented Aug 16, 2017

bterlson commented Aug 16, 2017

annevk commented Aug 16, 2017

bterlson commented Aug 17, 2017

Choose a reason for hiding this comment

bterlson commented Aug 17, 2017

Choose a reason for hiding this comment

domenic left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

bterlson commented Aug 17, 2017

Choose a reason for hiding this comment

bterlson commented Aug 17, 2017

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

annevk commented Aug 18, 2017

domenic commented Aug 18, 2017

Jack-Works commented May 12, 2020

claudepache commented May 12, 2020

Jack-Works commented May 12, 2020

claudepache commented May 12, 2020

claudepache Aug 19, 2016 •

edited

Loading

chicoxyzzy commented Aug 17, 2016 •

edited

Loading