API support for promises

API support for promises

R=svenpanne@chromium.org
BUG=

Committed: https://code.google.com/p/v8/source/detail?r=19811

[rossberg, 2014-03-11 10:56:08]

PTAL

[Sven Panne, 2014-03-11 11:44:05]

LGTM with nits

https://codereview.chromium.org/194663003/diff/1/include/v8.h
File include/v8.h (right):

https://codereview.chromium.org/194663003/diff/1/include/v8.h#newcode1417
include/v8.h:1417: * This is an experimental feature. Only works with --harmony
flag.
Hopefully this works without --harmony, too (and returns false in that case).

https://codereview.chromium.org/194663003/diff/1/src/contexts.h
File src/contexts.h (right):

https://codereview.chromium.org/194663003/diff/1/src/contexts.h#newcode176
src/contexts.h:176: V(PROMISE_CATCH_INDEX, JSFunction, promise_catch) \
Hmmm, I think this shouldn't make a difference, but anyway: Please run "make
grokdump" and see if it doesn't change tools/v8heapconst.py.

[yhirano, 2014-03-11 11:46:42]

https://codereview.chromium.org/194663003/diff/1/include/v8.h
File include/v8.h (right):

https://codereview.chromium.org/194663003/diff/1/include/v8.h#newcode2562
include/v8.h:2562: 
Can we have Then, too?

[rossberg, 2014-03-11 16:08:40]

https://codereview.chromium.org/194663003/diff/1/include/v8.h
File include/v8.h (right):

https://codereview.chromium.org/194663003/diff/1/include/v8.h#newcode1417
include/v8.h:1417: * This is an experimental feature. Only works with --harmony
flag.
On 2014/03/11 11:44:05, Sven Panne wrote:
> Hopefully this works without --harmony, too (and returns false in that case).

Done.

https://codereview.chromium.org/194663003/diff/1/include/v8.h#newcode2562
include/v8.h:2562: 
On 2014/03/11 11:46:42, yhirano wrote:
> Can we have Then, too?

Will that actually be useful on the API side? (I'm not even sure about
Chain/Catch. Let's keep the API minimal for now.)

https://codereview.chromium.org/194663003/diff/1/src/contexts.h
File src/contexts.h (right):

https://codereview.chromium.org/194663003/diff/1/src/contexts.h#newcode176
src/contexts.h:176: V(PROMISE_CATCH_INDEX, JSFunction, promise_catch) \
On 2014/03/11 11:44:05, Sven Panne wrote:
> Hmmm, I think this shouldn't make a difference, but anyway: Please run "make
> grokdump" and see if it doesn't change tools/v8heapconst.py.

Done.

[rossberg, 2014-03-11 16:17:28]

Committed patchset #2 manually as r19811 (presubmit successful).

[yhirano, 2014-03-12 01:23:15]

We have WebCore::ScriptPromise::then.
Though it is possible to implement it with general property access,
v8::Promise::Then is useful.

https://code.google.com/p/chromium/codesearch#chromium/src/third_party/WebKit...

[yhirano, 2014-03-14 05:57:00]

> Though it is possible to implement it with general property access,
I was wrong.
ScriptPromise.then should be immune against the environment modification by a
JavaScript developer.
Thus v8::Promise::Then is necessary.

[rossberg, 2014-03-14 07:14:32]

On 2014/03/12 01:23:15, yhirano wrote:
> We have WebCore::ScriptPromise::then.
> Though it is possible to implement it with general property access,
> v8::Promise::Then is useful.
> 
>
https://code.google.com/p/chromium/codesearch#chromium/src/third_party/WebKit...

Interesting. But why does this need Then instead of plain chaining? This is an
honest question. For example, I don't understand why you would need thenable
assimilation on the API side.

[yhirano, 2014-03-14 14:17:42]

On 2014/03/14 07:14:32, rossberg wrote:
> On 2014/03/12 01:23:15, yhirano wrote:
> > We have WebCore::ScriptPromise::then.
> > Though it is possible to implement it with general property access,
> > v8::Promise::Then is useful.
> > 
> >
>
https://code.google.com/p/chromium/codesearch#chromium/src/third_party/WebKit...
> 
> Interesting. But why does this need Then instead of plain chaining? This is an
> honest question. For example, I don't understand why you would need thenable
> assimilation on the API side.

ScriptPromise::then will be used by Blink developers to attach handlers to a
Promise provided by a JavaScript developer.
If we don't have |then|, we have to specify to pass a Promise that wraps an
object exactly once to DOM methods, i.e. a JavaScript developer should know if
it is a Promise of an object or a Promise of a Promise of an object. It is
nonsense I think. We need recursive unwrapping.