Ambiguity when `?.` in LHS?

[js-choi]

At the Committee plenary today, @waldemarhorwat expressed concern that (from my memory, might be incorrect) a.b?.c::fn is ambiguous.

I’m not sure where the ambiguity comes from yet. a.b?.c::fn is equivalent to (a.b?.c)::fn. The specification has an OptionalChain :: OptionalChain :: SimpleMemberExpression production, and a.b?.c::fn should be:

• OptionalExpression:
  • MemberExpression: a.b
  • OptionalChain:
    • OptionalChain:
      • ?.
      • IdentifierName: c
    • ::
    • SimpleMemberExpression: fn

But I may well be missing something.

[ljharb]

Perhaps, because to a human a.b?.c::fn() could mean:

1. (a.b?.c)::fn()
2. (a.b)?.(c::fn())

?

[js-choi]

@ljharb: Maybe, although I think that a.b?.c::fn() “obviously” reads as (a.b?.c)::fn(), in the same way that a.b?.c.d() “obviously” reads as (a.b?.c).d(), but maybe I’m wrong. I guess we’re not sure if @waldemarhorwat meant human visual ambiguity or grammatical ambiguity.

[waldemarhorwat]

Look at the meeting notes. The example I gave during the meeting is a?.b::c.d. The OptionalChain production parses it as both «a?.b::c».d and a?.b::«c.d».

[js-choi]

@waldemarhorwat: Thanks for the catch. This indeed seems like a bug. It seems like it should be fixable by changing:

OptionalChain :
  OptionalChain `::` SimpleMemberExpression

…to:

OptionalChain :
  OptionalChain `::` SimpleMemberExpression [lookahead != `.`]