Currently no way to define structs?

[waldemarhorwat]

The proposed spec does not currently provide any way to define structs or struct expressions — there's nothing I can find that references those productions from elsewhere in the grammar other than export.

I'm curious how syntactic collisions will be resolved once such expressions are defined.

[syg]

Oops, good catch on the omission. I was intending to add StructDeclaration as another production inside Declaration. Is struct [no new line terminator] insufficient to resolve the collisions?

Edit: and also StructExpression added as another production inside PrimaryExpression.

Also cc @rbuckton who works with the spec grammar much more than I do.

[waldemarhorwat]

You'll also have to resolve collisions between struct declarations and expressions.

[rbuckton]

For shared structs at least, there is no expression version as they must be top level declarations. I can't recall if we wanted to enforce that rule for non-shared structs as well, however.

[waldemarhorwat]

So you'd just drop the expression versions?

That would take care of the declaration-vs-expression conflict, as well as some other conflicts such as a struct expression whose name happens to be of.

[syg]

The expressions were added to the spec text more because I was copying class expression stuff from the main spec as a starting point than anything else. I think the utility of struct expressions is pretty low, so I've opened #38.

[acutmore]

I think the utility of struct expressions is pretty low

Was just chatting with @jridgewell who mentioned AMP (? if memory correct) has a transform to convert class declarations into class expressions to avoid TDZ checks.

Do structs follow suite, or maybe a chance to avoid TDZ checks?

[syg]

has a transform to convert class declarations into class expressions to avoid TDZ checks.

Interesting. I guess we'd have to weigh that against the grammar complexities for supporting expressions and declarations at the same time.

The correlation problem strongly favors restricting shared struct declarations to top-level only, which also implies no expressions.

As much as I hate TDZ it seems like a smaller issue here.