Allow void functions

[JustusAdam]

It has become apparent that it is necessary to support functions with no return. Perhaps we can abuse the unit hack again.

[sertel]

Here is the error report such that we can derive a test case:

This code has problems compiling:

ns some_ns;

use sf addition (produce, consume);

fn main() {
    let result = produce();
    consume(result)
}

It throws this error message:

[felix@Samaritan] $ ohuac build foobar.ohuac
[Debug] Lowering alang expr: AExpr {unAExpr = LetF "a" (AExpr {unAExpr = ApplyF (AExpr {unAExpr = VarF (Sf addition/consume Nothing)}) (AExpr {unAExpr = VarF (Sf addition/produce Nothing)})}) (AExpr {unAExpr = VarF (Local "a")})}
[Debug] Lowering Let -->
[Debug] Current state: []
[Error] Var must be local or env, was Sf addition/produce Nothing

The formatted version of the expression looks as follows:

AExpr {
  unAExpr = LetF "a" (AExpr { unAExpr = ApplyF (AExpr { unAExpr = VarF (Sf addition/consume Nothing)}) 
                                               (AExpr { unAExpr = VarF (Sf addition/produce Nothing)})}) 
                     (AExpr {unAExpr = VarF (Local "a")})}

According to this expression, the code actually looks as follows:

ns some_ns;

use sf addition (produce, consume);

fn main() {
    consume(produce());
}

In any case, the lowering pass seems to expect a Var as the second argument to Apply but finds Sf.
I already talked to @JustusAdam about this and this seems to be a problem with functions that do not have arguments.

[JustusAdam]

What you are talking about is ohua-dev/alang-clike-parser#7 which is already fixed. I will make a new release for ohuac soon which will include this fix.

What I mean by void functions is something like this

fn foo (a) {
    compute(a);
    flush();
}

Which I would like to desugar to

let foo = \a -> let _ = compute a in let _ = flush () in ()

[sertel]

(I was not sure whether it was already taken care of, so I added it to this issue.)

Sounds good to me.

[sertel]

@JustusAdam is still a valid work item? I know that I fixed this in core at least.

[JustusAdam]

I think so.