1270601 - Implement support for the proposed "namespace" IDL syntax

Status: RESOLVED FIXED

(Core :: DOM: Core & HTML, defect)

Description

[Boris Zbarsky [:bzbarsky]]

The idea is that this:

  namespace console {
    // stuff
  };

should act more or less like this:

  interface console {
    // static stuff
  };

with the following differences:

1)  typeof console === "object", not "function".  The call/construct hooks
    should be null.
2)  console.prototype === undefined.  I think we may already end up doing this
    for static-only stuff if we mark it abstract, fwiw.
3)  console.__proto__ == Object.prototype.

Comment 1

[Boris Zbarsky [:bzbarsky]]

Attachment: part 1. Add Web IDL parser support for IDL namespace syntax

Comment 2

[Boris Zbarsky [:bzbarsky]]

Attachment: part 2. Add codegen support for IDL namespaces

Comment 3

[Boris Zbarsky [:bzbarsky]]

Andrea, with those two patches you should be able to do:

  [ClassString="Console"]
  namespace console {
    ....
  };

Comment 4

[Boris Zbarsky [:bzbarsky]]

Attachment: Now with support for partial namespace too, per spec updates

Comment 5

[Boris Zbarsky [:bzbarsky]]

Attachment: And the updated codegen bits

Comment 6

[Andrea Marchesini [:baku]]

I need these patches in order to land bug 989619.

Comment 7

[Peter Van der Beken [:peterv]]

Comment on attachment 8752243 [details] [diff] [review]
Now with support for partial namespace too, per spec updates

Review of attachment 8752243 [details] [diff] [review]:
-----------------------------------------------------------------

::: dom/bindings/parser/WebIDL.py
@@ +800,5 @@
> +                    # Just mark all our methods/attributes as static.  The other
> +                    # option is to duplicate the relevant InterfaceMembers
> +                    # production bits but modified to produce static stuff to
> +                    # start with, but that sounds annoying.
> +                    m.forceStatic()

Would it make sense to do this in IDLNamespace's __init__?

@@ +1360,4 @@
>  
>      def addExtendedAttributes(self, attrs):
> +        # Note that this is only used by IDLInterface.  We could
> +        # consider moving this there...

Yeah, I think we should do that.

@@ +5437,5 @@
> +        if not nonPartialObject:
> +            nonPartialObject = nonPartialConstructor(
> +                # No members, False for isKnownNonPartial
> +                *(nonPartialConstructorArgs + [[], False]))
> +        partialInterface = IDLPartialInterface(*(partialConstructorArgs +

Might make sense to rename IDLPartialInterface (since it's now used for namespace too)?

::: dom/bindings/parser/tests/test_namespace.py
@@ +144,5 @@
> +        results = parser.finish()
> +    except Exception, x:
> +        threw = True
> +    harness.ok(threw, "Should have thrown.")
> +    

Trailing whitespace.

@@ +182,5 @@
> +        results = parser.finish()
> +    except Exception, x:
> +        threw = True
> +    harness.ok(threw, "Should have thrown.")
> +    

Trailing whitespace.

@@ +220,5 @@
> +        results = parser.finish()
> +    except Exception, x:
> +        threw = True
> +    harness.ok(threw, "Should have thrown.")
> +    

Trailing whitespace.

Comment 8

[Boris Zbarsky [:bzbarsky]]

Attachment: part 3. Add a way to annotate and IDL namespace as needing a new plain object as its prototype, because using Object.prototype as the prototype of 'console' is not web-compatible

Additional patch to deal with the compat issues described at https://github.com/whatwg/console/issues/3#issuecomment-219549472

Comment 9

[Boris Zbarsky [:bzbarsky]]

> Would it make sense to do this in IDLNamespace's __init__?

That makes partial stuff hard: the members of partials are just stored on those partials until we get to IDLInterfaceOrNamespace.finish.

> Yeah, I think we should do that.

Done.

> Might make sense to rename IDLPartialInterface (since it's now used for namespace too)?

Done, IDLPartialInterfaceOrNamespace.

Fixed the trailing whitespace.

Comment 10

[Peter Van der Beken [:peterv]]

Comment on attachment 8758313 [details] [diff] [review]
part 3.  Add a way to annotate and IDL namespace as needing a new plain object as its prototype, because using Object.prototype as the prototype of 'console' is not web-compatible

Review of attachment 8758313 [details] [diff] [review]:
-----------------------------------------------------------------

::: dom/bindings/BindingUtils.h
@@ +3274,5 @@
>  
> +namespace binding_detail {
> +inline
> +JSObject*
> +GetHackedNamespaceProtoObject(JSContext* aCx, JS::Handle<JSObject*> aForObj)

Can we not name the second argument, to make it clear it's unused?

Comment 11

[Boris Zbarsky [:bzbarsky]]

> Can we not name the second argument, to make it clear it's unused?

Done.

Comment 12

[Pulsebot]

Pushed by [e-mail]:
https://hg.mozilla.org/integration/mozilla-inbound/rev/9482e25918a3
part 1.  Add Web IDL parser support for IDL namespace syntax.  r=peterv
https://hg.mozilla.org/integration/mozilla-inbound/rev/1d761dc47c2d
part 2.  Add codegen support for IDL namespaces.  r=peterv
https://hg.mozilla.org/integration/mozilla-inbound/rev/508c43bf5578
part 3.  Add a way to annotate and IDL namespace as needing a new plain object as its prototype, because using Object.prototype as the prototype of 'console' is not web-compatible.  r=peterv

Comment 13

[Carsten Book [:Tomcat]]

https://hg.mozilla.org/mozilla-central/rev/9482e25918a3
https://hg.mozilla.org/mozilla-central/rev/1d761dc47c2d
https://hg.mozilla.org/mozilla-central/rev/508c43bf5578