Implement WebIDL implements

Implement WebIDL implements

Implement basic support for Web IDL implements statements in the
bindings generator:
http://www.w3.org/TR/WebIDL/#dfn-implements-statement

Introduce ParentNode interface and refactor Element, Document and
and DocumentFragment to use implements and match the spec:
http://dom.spec.whatwg.org/#parentnode

BUG=239236
Committed: https://src.chromium.org/viewvc/blink?view=rev&revision=152655

[do-not-use, 2013-06-18 14:59:39]

Note that with the current CL, we loose the [PerWorldBindings] optimization for
the Element attributes. I don't know how important this is.

Alternatively, we could use [PerWorldBindings] in ParentNode.idl and enable the
optimization for Element, Document and DocumentFragment.

[haraken, 2013-06-18 15:12:15]

Overall, this looks great! Let me take a detailed look tomorrow morning.

BTW, what's the difference between 'partial' and 'implements'? (Sorry, I'm
asking this without reading the spec.)

> Note that with the current CL, we loose the [PerWorldBindings] optimization
for
> the Element attributes. I don't know how important this is.
> 
> Alternatively, we could use [PerWorldBindings] in ParentNode.idl and enable
the
> optimization for Element, Document and DocumentFragment.

[PerWorldBindings] is a hint for performance optimization. With
[PerWorldBindings], we can drop one if branch in the generated code for the main
world. Ideally, we want to kill [PerWorldBindings] and generate two callbacks
for all DOM attributes and methods (one for the main world and the other for
isolated worlds). However, we realized that generating two callbacks for all DOM
attributes and methods unacceptably increases the binary size. So we decided to
generate two callbacks only for selected DOM attributes that are
performance-sensitive. They are marked as [PerWorldBindings]. In conclusion,
please don't drop the current [PerWorldBindings] :)

[do-not-use, 2013-06-18 15:16:45]

On 2013/06/18 15:12:15, haraken wrote:
> Overall, this looks great! Let me take a detailed look tomorrow morning.
> 
> BTW, what's the difference between 'partial' and 'implements'? (Sorry, I'm
> asking this without reading the spec.)
> 
> > Note that with the current CL, we loose the [PerWorldBindings] optimization
> for
> > the Element attributes. I don't know how important this is.
> > 
> > Alternatively, we could use [PerWorldBindings] in ParentNode.idl and enable
> the
> > optimization for Element, Document and DocumentFragment.
> 
> [PerWorldBindings] is a hint for performance optimization. With
> [PerWorldBindings], we can drop one if branch in the generated code for the
main
> world. Ideally, we want to kill [PerWorldBindings] and generate two callbacks
> for all DOM attributes and methods (one for the main world and the other for
> isolated worlds). However, we realized that generating two callbacks for all
DOM
> attributes and methods unacceptably increases the binary size. So we decided
to
> generate two callbacks only for selected DOM attributes that are
> performance-sensitive. They are marked as [PerWorldBindings]. In conclusion,
> please don't drop the current [PerWorldBindings] :)

Ok, I will add [PerWorldBindings] in ParentNode.idl then.

Regarding the difference, right now the only difference (with this CL) is that a
partial interface can only extend one interface. However, with implements
statements you can extend several (like in my CL).

[arv (Not doing code reviews), 2013-06-18 15:18:21]

https://codereview.chromium.org/17376006/diff/4001/Source/bindings/scripts/pr...
File Source/bindings/scripts/preprocess_idls.py (right):

https://codereview.chromium.org/17376006/diff/4001/Source/bindings/scripts/pr...
Source/bindings/scripts/preprocess_idls.py:161: # Parse 'identifier-A implements
identifier-B; statements
# Parse 'identifier-A implements identifier-B;' statements

https://codereview.chromium.org/17376006/diff/4001/Source/core/dom/ParentNode...
File Source/core/dom/ParentNode.idl (right):

https://codereview.chromium.org/17376006/diff/4001/Source/core/dom/ParentNode...
Source/core/dom/ParentNode.idl:4: readonly attribute HTMLCollection children;
Is there a problem if we have [PerWorldBindings] here?

The Element version of these had [PerWorldBindings] for speed.

[haraken, 2013-06-18 15:19:11]

> Regarding the difference, right now the only difference (with this CL) is that
a
> partial interface can only extend one interface. However, with implements
> statements you can extend several (like in my CL).

At a glance at the spec, 'partial' looks like just a subset of 'implements'. I'm
OK to have both in Blink because the spec has both, but I wonder why the spec
has both.

[arv (Not doing code reviews), 2013-06-18 15:19:31]

LGTM if you add back [PerWorldBindings]

[arv (Not doing code reviews), 2013-06-18 15:21:25]

On Tue, Jun 18, 2013 at 11:19 AM, <haraken@chromium.org> wrote:

> Regarding the difference, right now the only difference (with this CL) is
>> that
>>
> a
>
>> partial interface can only extend one interface. However, with implements
>> statements you can extend several (like in my CL).
>>
>
> At a glance at the spec, 'partial' looks like just a subset of
> 'implements'. I'm
> OK to have both in Blink because the spec has both, but I wonder why the
> spec
> has both.
>

partial is intended to allow a single interface object to be defined in
multiple places. This is similar to partial classes in C#.

-- 
erik

[do-not-use, 2013-06-18 15:22:01]

On 2013/06/18 15:19:31, arv wrote:
> LGTM if you add back [PerWorldBindings]

[PerWorldBindings] is back in patch set 3. One difference is that it now applies
to Document and DocumentFragment as well, not just Element.

[haraken, 2013-06-18 15:27:49]

Makes sense. LGTM2.

[commit-bot: I haz the power, 2013-06-18 17:05:35]

CQ is trying da patch. Follow status at
https://chromium-status.appspot.com/cq/ch.dumez@sisa.samsung.com/17376006/8002

[commit-bot: I haz the power, 2013-06-18 18:25:43]

Change committed as 152655