ES6: Array iterator methods

This adds the following array iterator methods:

Array.prototype.values
Array.prototype.keys
Array.prototype.entries

These all return an Array Iterator object which has a next
method.

http://people.mozilla.org/~jorendorff/es6-draft.html#sec-15.4.5

BUG=v8:2722

[arv (Not doing code reviews), 2013-06-12 21:45:32]

Feedback wanted.

This implements Array.prototype.{values,keys,entries}

It currently depends on Symbol for the private state. Is there an alternative
here to not have this dependency? It seems silly to have to implement the array
iterator in C++ to work around that.

[wingo, 2013-06-13 04:26:18]

Very cool!

Regarding the iterator state: why not use a closure instead?  I find publically
accessible internal state to be terrifying, but maybe that's just me :)

https://codereview.chromium.org/16848004/diff/1/src/array-iterator.js
File src/array-iterator.js (right):

https://codereview.chromium.org/16848004/diff/1/src/array-iterator.js#newcode36
src/array-iterator.js:36: var s = $Symbol();
Is this variable used?

[rossberg, 2013-06-13 08:21:47]

On 2013/06/12 21:45:32, arv wrote:
> Feedback wanted.
> 
> This implements Array.prototype.{values,keys,entries}
> 
> It currently depends on Symbol for the private state. Is there an alternative
> here to not have this dependency? It seems silly to have to implement the
array
> iterator in C++ to work around that.

Looks good so far.

Unfortunately, the only alternative to symbols seem to be weak maps, but they'd
be significantly more expensive. I honestly don't know what we should do about
this problem in general. I already feel tempted to add some non-standard API for
true private names/properties...

[rossberg, 2013-06-13 08:22:02]

https://codereview.chromium.org/16848004/diff/1/src/array-iterator.js
File src/array-iterator.js (right):

https://codereview.chromium.org/16848004/diff/1/src/array-iterator.js#newcode72
src/array-iterator.js:72: 
Nit: duplicated empty line

https://codereview.chromium.org/16848004/diff/1/test/mjsunit/harmony/array-it...
File test/mjsunit/harmony/array-iterator.js (right):

https://codereview.chromium.org/16848004/diff/1/test/mjsunit/harmony/array-it...
test/mjsunit/harmony/array-iterator.js:29: 
Can you add some tests using for-of?

[rossberg, 2013-06-13 08:23:27]

On 2013/06/13 04:26:18, wingo wrote:
> Regarding the iterator state: why not use a closure instead?

Unfortunately, that doesn't work, because the spec prescribes 'next' as a method
on the prototype. :(

[arv (Not doing code reviews), 2013-06-13 13:54:27]

Yeah, the spec requires next to be a method on the prototype so we need private
state.

Andreas, I'm in favor to exposing private names to embedders if we cannot get
TC39 to support them.

Is it OK to trigger this based on --harmony_iteration? Is it OK that this needs
--harmony_symbol to also be specified? Maybe a new flag would be cleaner do that
it doesn't fail when --harmony_symbols is not enabled?

https://codereview.chromium.org/16848004/diff/1/src/array-iterator.js
File src/array-iterator.js (right):

https://codereview.chromium.org/16848004/diff/1/src/array-iterator.js#newcode36
src/array-iterator.js:36: var s = $Symbol();
On 2013/06/13 04:26:18, wingo wrote:
> Is this variable used?

Done.

https://codereview.chromium.org/16848004/diff/1/src/array-iterator.js#newcode72
src/array-iterator.js:72: 
On 2013/06/13 08:22:02, rossberg wrote:
> Nit: duplicated empty line

Done.

https://codereview.chromium.org/16848004/diff/1/test/mjsunit/harmony/array-it...
File test/mjsunit/harmony/array-iterator.js (right):

https://codereview.chromium.org/16848004/diff/1/test/mjsunit/harmony/array-it...
test/mjsunit/harmony/array-iterator.js:29: 
On 2013/06/13 08:22:02, rossberg wrote:
> Can you add some tests using for-of?

I was planning to. I just wanted to send this out before I left the office
yesterday.

[arv (Not doing code reviews), 2013-07-08 22:03:16]

Sorry for letting this sit so long.

Cleaned up to not depend on --harmony-symbols. It now uses %CreateSymbol
directly.

The only question remaining is if we want a new flag for this
(harmony-array-iterator)?

[rossberg, 2013-07-09 13:03:39]

LGTM, will land

On 2013/07/08 22:03:16, arv wrote:
> Sorry for letting this sit so long.
> 
> Cleaned up to not depend on --harmony-symbols. It now uses %CreateSymbol
> directly.
> 
> The only question remaining is if we want a new flag for this
> (harmony-array-iterator)?

I think it's fine to just leave it under the harmony-iteration flag. It's
probably not very useful without for-of anyway.

[rossberg, 2013-07-11 11:32:19]

Committed, but that wasn't propagated here, because I hit Ctrl-C when I noticed
that I had forgotten the committer flag. But was too late already, sorry about
that!