Zig: Proposal: async prefix syntax

[andersfr]

The syntax specifies async<PrefixExpr> which is valid and non-ambiguous.

The problem is that it undermines parsing optimisations on expressions by terminating with a symbol that is also a legal infix operator (greater than). If you have seen c++ templates you understand the pain of parsing >.

Parsing strictly according to the grammar specification requires deep recursion on the most common constructs. This recursion/nesting can be optimised away but not if symbols are overloaded in this way.

I propose to use async(PrefixExpr) instead.

This is also consistent with how other function modifiers are used, e.g. align(8), linksection(".text")

[daurnimator]

The allocator is being removed from async very very soon. See the rewrite-coroutines branch.

[andersfr]

If the allocator disappears then this proposal becomes obsolete.

[andrewrk]

I opened the draft PR #3033 so that it's more visible