Using more widely used symbol polyfill or means of detecting `$iterator$`

[benlesh]

Currently this library uses a library-specific string for iterators if Symbol.iterator does not exist.

This has meant we need a strange workaround for RxJS to access iterators on types provided by es6-shim. This isn't ideal, because if you change your implementation of Set or Map, our workaround will break.

There are two fixes I'll propose for this, and neither are perfect:

1. Use a more common key for the iterator property
   • Basically use "@@iterator".
2. Publish the $iterator$ variable you're using somewhere globally so other libraries like RxJS can check for it and use it without workarounds.

[ljharb]

When we're in a browser that has a native property - like certain versions of Firefox with "@@iterator" - we use that. However, when we're not in a browser that has that, or Symbol, there's no global place to put it.

Where would you suggest?

Note that the "strange workaround" you're using is actually the perfect and correct approach to reliably get this value - I'm not sure why you'd need a different one.

[benlesh]

Note that the "strange workaround" you're using is actually the perfect and correct approach to reliably get this value - I'm not sure why you'd need a different one.

If you ever change your implementation of Map or Set, such that Map.prototype[$yourIterator$] !== Map.prototype.entries, then the workaround breaks.

[ljharb]

@blesh if we did that, we'd be violating the spec - thus, we will never do that.

[ljharb]

To clarify, the spec requires that Map.prototype[Symbol.iterator] === Map.prototype.entries && Set.prototype[Symbol.iterator] === Set.prototype.values && Array.prototype[Symbol.iterator] === Array.prototype.values be true.

[benlesh]

Interesting. Thanks. I couldn't locate the spec on this.

It's still unfortunate that we have a special case for es6-shim due to a somewhat avant-garde symbol property name. (@@iterator would have been nice)

Thanks for your feedback.

[ljharb]

I explicitly chose not to use "@@iterator" to avoid people wrongly attempting to "polyfill" iterators based on Firefox's unfortunate choice to make a nonstandard string property magical.

[benlesh]

I think we all would have been better off if people had just stuck to a few agreed upon conventions, honestly. But it's no big deal. It'll all be gone in a few years.

[ljharb]

I love your optimism :-) the first browser that didn't need the es5-shim appeared within the last year, 6 years after ES5 was standard.

[benlesh]

Haha... Aw, 6 years isn't a big deal. Look how long it took Java to get generics or arrow functions, and that's just one platform.