Introduce Oxpecker.ModelValidation

[Lanayx]

Validation model based on standard ASP.NET Core validation mechanism based on System.ComponentModel.DataAnnotations attributes.

[habib-sadullaev]

as an option

Suggested change

	match dict.TryGetValue(memberName) with
	| true, value -> value.Add(error.ErrorMessage)
	| false, _ ->
	let arrayList = ResizeArray(1)
	arrayList.Add(error.ErrorMessage)
	dict[memberName] <- arrayList
	if not (dict.Contains(memberName)) then dict[memberName] <- ResizeArray(1)
	dict[memberName].Add(error.ErrorMessage)

[Lanayx]

This will cause more dict access

[xperiandri]

Maybe NotValidated to prevent any collision.
However, if you require qualified access this is fine too

[Lanayx]

Yes, RequireQualifiedAccess resolves any collisions

[habib-sadullaev]

Suggested change

	member this.Value(f: 'T -> string | null) =
	match this with
	| Empty -> null
	| Valid model -> f model
	| Invalid(model, _) -> f model
	member this.BoolValue(f: 'T -> bool) =
	match this with
	| Empty -> false
	| Valid model -> f model
	| Invalid(model, _) -> f model
	private member this.Apply(f: 'T -> 'U) =
	match this with
	| Empty -> Unchecked.defaultof<'U>
	| Valid model -> f model
	| Invalid(model, _) -> f model
	member this.Value(f: 'T -> string | null) = this.Apply f
	member this.BoolValue(f: 'T -> bool) = this.Apply f

[Lanayx]

While I get the duplication point, I like the readability of the current implementation more (and it doesn't use Unchecked module)

[habib-sadullaev]

Do we really need to do this nullable?

[Lanayx]

Yes

[AP-Hunt]

Would this be better/more idiomatic as an Option<string>? Curious why the nullability here. I don't write F# in my day job, so I might be missing some nuance.

[Lanayx]

Right, the nuance is that it should be mostly used to set value attribute of the input and that has type string | null for performance and DX reasons. So I tried to match types here to avoid unnecessary conversions.

[AP-Hunt]

Oh that makes a lot of sense! Thanks for the explanation