Review notes

[anba]

https://ljharb.github.io/String.prototype.matchAll/#String.prototype.matchAll

Missing ReturnIfAbrupt after step 1. (not required if steps 5-6 are placed directly after step 1)

Steps 5-6 should be placed directly after step 1.

Step 10: Calling RegExpCreate with regexp as the pattern parameter is probably not intended. This either needs to be changed to:

1. Let pattern be ToString(Get(regexp, "source")).
2. ReturnIfAbrupt(pattern).
3. Let rx be RegExpCreate(pattern, flags).

Or alternatively Symbol.species should be used like in RegExp.prototype[Symbol.split] (http://tc39.github.io/ecma262/#sec-regexp.prototype-@@split, steps 5-6 and step 13). I guess Symbol.species is actually the correct choice here, even if Symbol.matchAll doesn't get added.

Nit: Missing full stop in step 15.

Note 2 is not correct, e.g. rx is exposed to user script if RegExp.prototype.exec is overridden.

https://ljharb.github.io/String.prototype.matchAll/#CreateRegExpStringIterator

It's not valid to call IsRegExp and Get in steps 1-2 because it can trigger side effects if RegExp.prototype[Symbol.match] or RegExp.prototype.global are not the intrinsic functions.

https://ljharb.github.io/String.prototype.matchAll/#%RegExpStringIteratorPrototype%.next

Missing ReturnIfAbrupt after step 6 (modified RegExp.prototype.exec!).

https://ljharb.github.io/String.prototype.matchAll/#PropertiesOfRegExpStringIteratorInstances

"IsRegExp([[IteratingRegExp]]) is always true." is not correct, same reason as above.

[ljharb]

Thanks for the thorough review! I still need to update the spec text after the latest meeting - specifically, to a) accept either a string or a regex, and b) to create Symbol.matchAll so that subclassing is handled properly. I'll update this issue when I've done so.