Why not allow await?

[bakkot]

It's surprising to me that this forbids await. I'd expect to be able to do anything I could do in the surrounding code, including initialization of the class itself which might require waiting for some asynchronous operation.

Note that you can already await in computed property names:

(async () => class {
  [await 0](){}
})()

is legal today.

This can't just be legalized in a followup because it interacts with parsing: if the await flag is not passed to the static block, then

class A {
  static {
    console.log(await);
  }
}

is syntactically legal even in an async context. To keep the possibility of legalizing this open, at the very least we'd need to pass the correct grammar flag today and just make AwaitExpression an error. But it seems both simpler and more useful to inherit async-ness from the surrounding context.

(Edit: I should mention that similar concerns apply to yield, though the parsing isn't as tricky.)

[ljharb]

Wouldn't allowing it mean that the partially-constructed class would be observable for more than just the rest of the static block?

[bakkot]

How so? Code after the class wouldn't run until the await completed, and code in other functions wouldn't have visibility of the class.

[ljharb]

What I mean is, the class could provide its partially-constructed class to other code, and then finish with the static block code after N ticks had passed and the class had been mutated. I'm not sure why that seems worse to me than the exact same thing happening synchronously, though, so maybe it's fine.

[rbuckton]

This is essentially a duplicate of #7 and you can read my rationale for why this is currently disallowed there. We may need to do something more to disallow await as an IdentifierReference inside of a static {} block to ensure we future-proof it should we decide to allow it later.

There shouldn't be a concern for yield, as yield is treated as a reserved word in strict mode so its already disallowed as an identifier inside of a class body. Its unfortunate await isn't treated as a reserved word in strict mode as well.

[rbuckton]

I'm closing this as a duplicate of #7. I've opened #27 to discuss the restrictions needed to both disallow await but leave it as an option for a follow-on proposal.

[bakkot]

#7 doesn't really include any rationale, that I can see. It just says it's not permitted.