built-in iterators should be disposable

[rbuckton]

This updates the definitions for IteratorObject and AsyncIteratorObject to be disposable, per the spec:

• https://arai-a.github.io/ecma262-compare/?pr=3000&id=sec-%25iteratorprototype%25-%40%40dispose
• https://arai-a.github.io/ecma262-compare/?pr=3000&id=sec-%25asynciteratorprototype%25-%40%40asyncdispose

It should be noted that %IteratorPrototype% is synonymous with Iterator.prototype per the Iterator Helpers proposal and is represented by IteratorObject in TypeScript, and that %AsyncIteratorPrototype% is synonymous with AsyncIterator.prototype per the Async Iterator Helpers proposal and is represented by AsyncIteratorObject in TypeScript.

Fixes #59263

[sandersn]

I like the code change, but I had a couple of questions about the tests.

[sandersn]

I don't understand why these succeed when using it7 = i fails. They're all Iterators too. Where do they get a [Symbol.dispose] method?

[rbuckton]

it7 is illustrating that the Iterator interface (which only describes a minimal Iterator compatible with for..of) does not implement Symbol.dispose. Only iterators that inherit from the IteratorObject interface (which represents the global Iterator.prototype value) will have a Symbol.dispose by default.

Iterators returned from values()/keys()/entries() on Map/Set/Array/etc. will inherit from IteratorObject and thus are disposable.

[sandersn]

Are these tests of existing functionality? I don't see how changing the type of [Async]IteratorObject would change its runtime behaviour.

[rbuckton]

This test was in the wrong test suite. I merely moved while adding other tests.

[trevorade]

I'm seeing some build errors that are related to this:

Rather than,

interface IteratorObject<T, TReturn, TNext> extends Disposable {
}

Should it have been:

interface Iterator<T, TReturn, TNext> extends Disposable {
}

IteratorObject extends Iterator so the mismatch of IteratorObject being disposable but Iterator not being disposable causes compiler errors in other locations.

[jakebailey]

No, it being on the interface isn't correct, I think you just need to update your @types/node to pick up fixes we made (specifically, removing the redeclaration of entries).

[trevorade]

Thx for the guidance!

For anyone else looking into this, the referenced PR is: DefinitelyTyped/DefinitelyTyped@8c36e53