Revamp interface bindings #283

tobie · 2017-01-24T23:59:58Z

Extract legacy callback interface objects
out of interface objects. Closes Legacy callback interface objects are badly defined #78.
Clarify that legacy callback interfaces are function objects.
Give them a length property. Closes Should callback interface objects have a "length" property? #279. Closes Check if NodeFilter.length is defined correctly. #83.
Require new for Named Constructors. Closes Require new for named constructors #275.
Rely on ES abstract operations for defining,
named constructors, interface objects, and
legacy callback interface objects.
Disallow using the [NoInterfaceObject] on
callback interfaces with constants.
Don't require an interface prototype object for constants.
Callback interfaces do not have an interface prototype object.
Their constants sit on the legacy callback interface object itself.
Closes Constants require an interface prototype object which callback interfaces don't have #281.

tobie · 2017-01-25T00:09:23Z

index.bs

 that is the value of the aforementioned property.

-The identifier used for the named constructor must not
+The [=NamedConstructor identifier|identifier=] used for the named constructor must not


I'm not sure this is really useful.

I could go either way. We don't have separate #x-identifier concepts for everything else. On the other hand, the "after the = sign" thing down below does sure seem out of place, so moving it up here is not a bad idea.

tobie · 2017-01-25T00:21:35Z

I'm not sure how ready this thing is for review. Will need to read through it tomorrow (or more likely, Thursday).

foolip · 2017-01-25T08:02:08Z

It's probably fine to disallow using [NoInterfaceObject] on callback interfaces period. At least I wouldn't expect to see any new callback interfaces on the platform, and the existing ones don't have have [NoInterfaceObject].

domenic

This is really, really great modernization and bug-fixing work. Found some nits, but overall quite happy.

domenic · 2017-01-26T19:30:04Z

index.bs

 that is the value of the aforementioned property.

-The identifier used for the named constructor must not
+The [=NamedConstructor identifier|identifier=] used for the named constructor must not


I could go either way. We don't have separate #x-identifier concepts for everything else. On the other hand, the "after the = sign" thing down below does sure seem out of place, so moving it up here is not a bad idea.

domenic · 2017-01-26T19:34:53Z

index.bs

+            then [=ECMAScript/throw=] a <emu-val>TypeError</emu-val>.
+        1.  If [=NewTarget=] is <emu-val>undefined</emu-val>, then
+            [=ECMAScript/throw=] a <emu-val>TypeError</emu-val>.
+        1.  Let |arg|<sub>0..|n|−1</sub> be arguments.


"the passed arguments" (here and below)

domenic · 2017-01-26T19:36:06Z

index.bs

+        1.  Return |O|.
+    1.  Let |proto| be the [=%FunctionPrototype%=] of |realm|.
+    1.  If |I| inherits from some other interface |P|,
+        then set |proto| to the [=interface object=] of |P|.


"interface object of P in realm".

domenic · 2017-01-26T19:36:17Z

index.bs

-    Otherwise, the value is determined as follows:
+    The [=interface object=]
+    for a given non-callback [=interface=] |I| with [=identifier=] |id|
+    and [=Realm=] |realm| is created as follows:


"and in Realm realm"

domenic · 2017-01-26T19:37:18Z

index.bs

+        PropertyDescriptor{\[[Value]]: |length|, \[[Writable]]: <emu-val>false</emu-val>, \[[Enumerable]]: <emu-val>false</emu-val>, \[[Configurable]]: <emu-val>true</emu-val>}).
+    1.  Let |obj| be the [=interface prototype object=] of [=interface=] |I|.
+    1.  Perform [=!=] [=DefinePropertyOrThrow=](|F|, "prototype",
+        PropertyDescriptor{\[[Value]]: |obj|, \[[Writable]]: <emu-val>false</emu-val>, \[[Enumerable]]: <emu-val>false</emu-val>, \[[Configurable]]: <emu-val>false</emu-val>}).


s/obj/proto; surprised the var checker didn't catch this

domenic · 2017-01-26T19:38:57Z

index.bs

-This object also must be
-associated with the ECMAScript global environment associated
-with the [=named constructor=].
+If the internal \[[Call]] method of the [=named constructor=] returns normally,


Prexisting problem, but: this shouldn't be talking about the internal [[Call]] method. It should be talking about the steps that spec authors use to describe the named constructor ("the actions listed in the description of constructor" below).

domenic · 2017-01-26T19:39:32Z

index.bs

@@ -10141,87 +10093,67 @@ which allows construction of objects that
 implement the interface on which the
 [{{NamedConstructor}}] extended attributes appear.


I would remove the "must have a [[Call]] property" sentence; that is implicit and restating it is confusing (how could it not be true)?

domenic · 2017-01-26T19:40:52Z

index.bs

-    A named constructor must have a property named “length” with attributes
-    { \[[Writable]]: <emu-val>false</emu-val>, \[[Enumerable]]: <emu-val>false</emu-val>, \[[Configurable]]: <emu-val>true</emu-val> }
-    whose value is a <emu-val>Number</emu-val> determined as follows:
+    Assuming |id| is the [=NamedConstructor identifier|identifier=] of the constructor


I'd restate this intro as you did for interface objects. E.g.

The named constructor for a given non-callback interface I with identifier id in Realm realm is created as follows:

domenic · 2017-01-26T19:43:38Z

index.bs

+then the interface prototype object must also
+have a property named “constructor” with attributes
+{ \[[Writable]]: <emu-val>true</emu-val>, \[[Enumerable]]: <emu-val>false</emu-val>, \[[Configurable]]: <emu-val>true</emu-val> }
+whose value is a reference to the interface object for the interface.


Link interface object while you're here?

domenic · 2017-01-26T19:45:29Z

index.bs

+    and [=Realm=] |realm| is created as follows:
+
+    1.  Let |steps| be the following steps:
+        1.  [=ECMAScript/throw=] a <emu-val>TypeError</emu-val>.


Capitalize throw

tobie · 2017-02-01T11:56:51Z

I'll rebase when we're done with the review (I think that's screws up the review process the least).

domenic

A few more nits, but feel free to merge after fixing them.

domenic · 2017-02-03T19:34:00Z

index.bs


-<h5 id="es-constructible-interfaces" oldids="es-interface-call">Constructible Interfaces</h5>
+The [=interface object=] for a given [=interface=] is a [=function object=]


Missing period

domenic · 2017-02-03T19:37:48Z

index.bs

-associated with the ECMAScript global environment associated
-with the [=named constructor=].
+If the actions listed in the description of the constructor return normally,
+then the [=named constructor=] must return an object that implements interface |I|.


"then those steps must return".

domenic · 2017-02-03T19:39:07Z

index.bs

+If the actions listed in the description of the constructor return normally,
+then the [=named constructor=] must return an object that implements interface |I|.
+This object also must be associated with the ECMAScript global environment
+associated with the [=named constructor=].


"This object's relevant Realm must be the same as that of the named constructor"

domenic · 2017-02-03T19:41:34Z

index.bs

+and its value is an object called the
+<dfn id="dfn-legacy-callback-interface-object" export>legacy callback interface object</dfn>.
+The property has the attributes { \[[Writable]]: <emu-val>true</emu-val>, \[[Enumerable]]: <emu-val>false</emu-val>, \[[Configurable]]: <emu-val>true</emu-val> }.
+The characteristics of a legacy callback interface object are described in [[#legacy-callback-interface-object]].


This last sentence is just a link to the section that the sentence is already in, which seems bad.

* Extract legacy callback interface objects out of interface objects. Closes whatwg#78. * Clarify that legacy callback interfaces are function objects. * Give them a length property. Closes whatwg#279. Closes whatwg#83. * Require new for Named Constructors. Closes whatwg#275. * Rely on ES abstract operations for defining, named constructors, interface objects, and legacy callback interface objects. * Disallow using the [NoInterfaceObject] on callback interfaces with constants.

Callback interfaces do not have an interface prototype object. Their constants sit on the legacy callback interface object itself. Closes whatwg#281.

* Extract legacy callback interface objects out of interface objects. Closes #78. * Clarify that legacy callback interfaces are function objects. * Give them a length property. Closes #279. Closes #83. * Require new for Named Constructors. Closes #275. * Rely on ES abstract operations for defining named constructors, interface objects, and legacy callback interface objects. * Disallow using [NoInterfaceObject] on callback interfaces with constants. * Don't require an interface prototype object for constants. Callback interfaces do not have an interface prototype object. Their constants sit on the legacy callback interface object itself. Closes #281. Closes #283.

tobie commented Jan 25, 2017

View reviewed changes

domenic reviewed Jan 26, 2017

View reviewed changes

tobie mentioned this pull request Feb 2, 2017

Clarify [[Call]] + [[Construct]] for named (and possibly non-named) constructors #278

Closed

domenic approved these changes Feb 3, 2017

View reviewed changes

domenic and others added 4 commits February 14, 2017 10:02

Set up auto-deploy to GitHub pages from master (whatwg#221)

e90a81a

Editorial: fix readme line breaks (whatwg#310)

84b6c30

Meta: fix broken deploy script (whatwg#311)

28c5514

Meta: add .pr-preview.json file.

535ac78

tobie force-pushed the interface-objs branch from 6184b2c to 83e0087 Compare February 20, 2017 21:47

tobie mentioned this pull request Feb 20, 2017

Revamp interface bindings #313

Merged

tobie added 6 commits February 20, 2017 23:05

Meta: add index.html to .gitignore

9830a84

Address review comments.

26e657e

Add advisement for NamedConstructors.

3a03cf4

Don't require an interface prototype object for constants.

4ecbd80

Callback interfaces do not have an interface prototype object. Their constants sit on the legacy callback interface object itself. Closes whatwg#281.

Address review comments.

dd73cf7

tobie force-pushed the interface-objs branch from b7c5579 to dd73cf7 Compare February 20, 2017 22:11

tobie closed this in #313 Feb 20, 2017

Provide feedback

Saved searches

Use saved searches to filter your results more quickly

Revamp interface bindings #283

Revamp interface bindings #283

tobie commented Jan 24, 2017 •

edited

Loading

tobie Jan 25, 2017

domenic Jan 26, 2017

tobie commented Jan 25, 2017

foolip commented Jan 25, 2017

domenic left a comment

domenic Jan 26, 2017

domenic Jan 26, 2017

domenic Jan 26, 2017

domenic Jan 26, 2017

domenic Jan 26, 2017

tobie Feb 1, 2017 •

edited

Loading

domenic Jan 26, 2017

domenic Jan 26, 2017

domenic Jan 26, 2017

domenic Jan 26, 2017

domenic Jan 26, 2017

tobie commented Feb 1, 2017

domenic left a comment

domenic Feb 3, 2017

domenic Feb 3, 2017

domenic Feb 3, 2017

domenic Feb 3, 2017

		@@ -10141,87 +10093,67 @@ which allows construction of objects that
		implement the interface on which the
		[{{NamedConstructor}}] extended attributes appear.


		<h5 id="es-constructible-interfaces" oldids="es-interface-call">Constructible Interfaces</h5>
		The [=interface object=] for a given [=interface=] is a [=function object=]

Revamp interface bindings #283

Revamp interface bindings #283

Conversation

tobie commented Jan 24, 2017 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

tobie commented Jan 25, 2017

foolip commented Jan 25, 2017

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

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

tobie Feb 1, 2017 • 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

tobie commented Feb 1, 2017

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

Choose a reason for hiding this comment

Choose a reason for hiding this comment

tobie commented Jan 24, 2017 •

edited

Loading

tobie Feb 1, 2017 •

edited

Loading