Splitting ExperimentalStroustrupStyle to separate settings

[Thorium]

I love the ExperimentalStroustrupStyle setting in records, but I hate it in computation expressions.
Computation expressions are so crucial part of F# that it should be very clear what goes on there.

What we want to optimize? Maintenance.
Let's observe the following code:

let myMethod() =
    async {
        let! x =
           async {
               do! waitSomething()
               return 4
           }
        return {
            Id = 1
            Data1 = "asdf"
            Data2 = "asdf"
            Data3 = x
        }
    }

There is no way to create this properly on any settings in Fantomas.

[nojaf]

Hello, thank you for opening this discussion on the Fantomas repository.

Splitting ExperimentalStroustrupStyle is an interesting case.

fantomas/src/Fantomas.Core/SourceParser.fs

Lines 1819 to 1831 in 434df47

	let (|StroustrupStyleExpr|_|) (isStroustrupStyleEnabled: bool) (e: SynExpr) =
	if not isStroustrupStyleEnabled then
	None
	else
	match e with
	// { foo with Bar = bar }
	| SynExpr.Record(copyInfo = Some _) -> None
	| SynExpr.Record _
	| SynExpr.AnonRecd _
	// task { ... }
	| SynExpr.App (ExprAtomicFlag.NonAtomic, false, SynExpr.Ident _, SynExpr.ComputationExpr _, _)
	| ArrayOrList _ -> Some e
	| _ -> None

Stroupstrous current works on records, array/list and computation expressions.
You can imagine that if we split one thing, people will eventually ask for everything to be split.
This will bring in an additional churn to maintain Fantomas.
It might be worth having, I would like to see what the community thinks about this.
Are enough people interested in this to carry it?
And of course, what big of an impact it will have on the codebase. Because there would be more changes of course than just the snippet above.

[josh-degraw]

Is your issue here essentially that you want async { to be on its own line, rather than on the same line as the binding? Because if so, in my opinion that's not (or shouldn't be) directly related to the Stroustrup style, which is mainly just about where the braces are placed relative to identifiers. IMO your example still conforms to Stroustrup style with the brackets, so I'd say it would make sense to allow that, but as a different flag, like compututation_expressions_on_newline or something like that.

[nojaf]

Available in v6.