A code refactoring to convert bool to a Yes|No DU

[psfinaki]

Update: actually turns out there are more ideas. See my original idea below and even more original ideas of kerams even more below.

---

Convert

type T = T of bool

to

[<RequireQualifiedAccess>]
type T = 
   | Yes
   | No

A true functional fix :)

Originally posted by @kerams in #15561 (comment)

[kerams]

No, no, I meant

let CheckIWSAM (cenv: cenv) (env: TcEnv) checkConstraints (checkIwsam: bool) m tcref =
    let g = cenv.g
    let ty = generalizedTyconRef g tcref

    if checkIwsam && isInterfaceTy g ty && checkConstraints = CheckCxs then
        let meths = AllMethInfosOfTypeInScope ResultCollectionSettings.AllResults cenv.infoReader env.NameEnv None env.eAccessRights IgnoreOverrides m ty
        ()

let TcLongIdentType kindOpt (cenv: cenv) newOk checkConstraints occ iwsam env tpenv synLongId =
    let (SynLongIdent(tc, _, _)) = synLongId
    let m = synLongId.Range
    let ad = env.eAccessRights

    CheckIWSAM cenv env checkConstraints true m tcref

->

[<RequireQualifiedAccess>]
type CheckIwsam =
    | Yes
    | No

let CheckIWSAM (cenv: cenv) (env: TcEnv) checkConstraints (checkIwsam: CheckIwsam) m tcref =
    let g = cenv.g
    let ty = generalizedTyconRef g tcref

    if checkIwsam = CheckIwsam.Yes && isInterfaceTy g ty && checkConstraints = CheckCxs then
        let meths = AllMethInfosOfTypeInScope ResultCollectionSettings.AllResults cenv.infoReader env.NameEnv None env.eAccessRights IgnoreOverrides m ty
        ()

let TcLongIdentType kindOpt (cenv: cenv) newOk checkConstraints occ iwsam env tpenv synLongId =
    let (SynLongIdent(tc, _, _)) = synLongId
    let m = synLongId.Range
    let ad = env.eAccessRights

    CheckIWSAM cenv env checkConstraints CheckIwsam.Yes m tcref

This is obviously much more complicated as we'd need to rewrite the definition site and all call sites.

[vzarytovskii]

Convert

type T = T of bool

to

[<RequireQualifiedAccess>]
type T = 
   | Yes
   | No

A true functional fix :)

Originally posted by @kerams in #15561 (comment)

Thats not very useful. It will be much more useful to remove boolean parameter and use newly introduced DU though.

However, it's a common pattern in compiler, but I rarely see it elsewhere, so not sure about overall usefulness.

Update: @kerams beat me to it.

[psfinaki]

Ohhhh I see. Alright, sorry for the confusion then. :)

Well I personally generally dislike whatever booleans wherever since they hinder readability on a calling side (bold statement I know). So probably both ideas could be worthy.

I'd wait for more people to chime in here with their opinions before splitting the ticket and jumping on it. Anyway, thanks for good examples up there.