MapIterator.next definition in spec

The latest spec draft (feb 2) still lists MapIterator.next as
returning an Array instance containing two elements. The
implementation of MapIterator in latest SpiderMonkey and V8 returns a
JS object, like { value: x, done: false }.

Is the current spec draft correct? Do applications using Map have to
detect which of the two formats is being used by the current iterator
implementation?

Most of the stuff out there on the web (MDN, caniuse, etc) suggests
that the implementations in V8 and SpiderMonkey are current to the
draft spec and reasonably complete, so this is quite confusing.

Thanks,
-kg

> Le 4 févr. 2015 à 08:57, Katelyn Gadd <kg at luminance.org> a écrit :
> 
> The latest spec draft (feb 2) still lists MapIterator.next as
> returning an Array instance containing two elements. The
> implementation of MapIterator in latest SpiderMonkey and V8 returns a
> JS object, like { value: x, done: false }.

If I read correctly the algorithm in Section 23.1.5.2.1 %MapIteratorPrototype%.next(), 
MapIterator#next() returns either `CreateIterResultObject(result, false)`,
or `CreateIterResultObject(undefined, true)` (when enumeration is completed),
where `CreateIterResult(result, done)` is `{value: result, done: done}`.

Is there something in another part of the spec contradicting that?

—Claude

Here's what I see in the feb 2 pdf:

1. Assert: itemKind is "key+value".
2. Let result be the result of performing ArrayCreate(2).
3. Assert: result is a new, well-formed Array object so the following
operations will never fail.
4. Call CreateDataProperty(result, "0", e.[[key]]) .
5. Call CreateDataProperty(result, "1", e.[[value]]).

On 4 February 2015 at 00:23, Claude Pache <claude.pache at gmail.com> wrote:
>
>> Le 4 févr. 2015 à 08:57, Katelyn Gadd <kg at luminance.org> a écrit :
>>
>> The latest spec draft (feb 2) still lists MapIterator.next as
>> returning an Array instance containing two elements. The
>> implementation of MapIterator in latest SpiderMonkey and V8 returns a
>> JS object, like { value: x, done: false }.
>
> If I read correctly the algorithm in Section 23.1.5.2.1 %MapIteratorPrototype%.next(),
> MapIterator#next() returns either `CreateIterResultObject(result, false)`,
> or `CreateIterResultObject(undefined, true)` (when enumeration is completed),
> where `CreateIterResult(result, done)` is `{value: result, done: done}`.
>
> Is there something in another part of the spec contradicting that?
>
> —Claude
>

> Le 4 févr. 2015 à 10:02, Katelyn Gadd <kg at luminance.org> a écrit :
> 
> Here's what I see in the feb 2 pdf:
> 
> 1. Assert: itemKind is "key+value".
> 2. Let result be the result of performing ArrayCreate(2).
> 3. Assert: result is a new, well-formed Array object so the following
> operations will never fail.
> 4. Call CreateDataProperty(result, "0", e.[[key]]) .
> 5. Call CreateDataProperty(result, "1", e.[[value]]).

So do I; but the algorithm doesn't say "Return result". The next step is:

  iv. Return CreateIterResultObject(result, false).

—Claude

I see, that's why I got confused, since that's defined elsewhere in
the spec. Makes sense then, sorry for the error.

On 4 February 2015 at 01:37, Claude Pache <claude.pache at gmail.com> wrote:
>
>> Le 4 févr. 2015 à 10:02, Katelyn Gadd <kg at luminance.org> a écrit :
>>
>> Here's what I see in the feb 2 pdf:
>>
>> 1. Assert: itemKind is "key+value".
>> 2. Let result be the result of performing ArrayCreate(2).
>> 3. Assert: result is a new, well-formed Array object so the following
>> operations will never fail.
>> 4. Call CreateDataProperty(result, "0", e.[[key]]) .
>> 5. Call CreateDataProperty(result, "1", e.[[value]]).
>
> So do I; but the algorithm doesn't say "Return result". The next step is:
>
>   iv. Return CreateIterResultObject(result, false).
>
> —Claude
>
>