Ast async

[cristianoc]

Prepare for ast async cleanup:

• Refactor code for "@res.async" payload handling
• Clean up handling of type and term parameters: now each => in a function definition corresponds to a function, even when only types are present. So e.g. (type a) => 3 is now a function just like () => 3.

The question remains on what to do with trailing types in function definitions: (x, type a) => x+1. There's no way currently to represent them cleanly. One could forbid these case or move all the type a before all the term parameters.

[zth]

Awesome! Is there any difference in practice between a prefixed or trailing term type?

[cknitt]

Tested against a real world project without issues. 👍

[cristianoc]

Awesome! Is there any difference in practice between a prefixed or trailing term type?

Technically, mixing types and terms provides some restrictions (terms appearing before a type cannot use it). For example, this cannot be typed:

(x, type a, y:a) => x == y

That said, I could not find examples where this restriction is actually useful.
Thinking about having a normalised form (type a b c, x, y, z) => ... where all the types appear at the beginning. Which is the shape that all existing examples take.

In fact, I'm considering to treat types as if they were all in the beginning anyway, and reformatting to that form.
One additional side effect of this form, is that we could eventually remove type a as an independent AST node, and instead move all types to the AST node for function definitions. Eventually, (type a b c, x, y, z) could be a single node.