Pre SIP: `for` with Control Flow (an alternative to Scala Async, Scala Continuations and Scala Virtualized)

[yangbo]

@rssh

• in match I afraid about nesting .right expression ‘k-times’.

Why?

@rssh

• block part looks unfinished or too restrictive (not specified, how to handle ← expression inside stats ).

That should follow the existing for comprehension translation rules, since the proposed rules are just some addition to existing rules.

@rssh

• next steps having x ← y instead x = await(x) wrapped in some async expression: the value for developer will be the value of avoiding àsync brackets on top of each function.

That’s exact the approach that has been implemented in Dsl.scala’s ResetEverywhere plugin. It might be a good idea.

@rssh

– a) if we want ‘lift’ type of monad to the type of DSL expression, then we would have untrivial inference with typing;

I would suggest to use Dsl type class instead of Monad, because Dsl has been proven it works well with Scala type inference, even in complex use cases.

@rssh

– b) proposal as specified less powerful than existing async wrappers, where while -s and nested blocks are supported.

Why do you say that? I think this proposal also addresses while and nested blocks.

@rssh

• also specifying concrete rules in specification prevent optimizations (for example, dotty-cps-async have different patterns when all subexpression are async or some not, so we have no more than minimum nesting of monads applications. )

I disagree. Virtualized control flow provides the ability of optimization in runtime.