Broken equality between enum and union of its values

[nin-jin]

Bug Report

🕗 Version & Regression Information

• This changed between versions 4.3.5 and 4.4.3

💻 Code

type Equals< A , B > =
		(
			<X>()=> X extends A ? 1 : 2
		) extends (
			<X>()=> X extends B ? 1 : 2
		)
			? true
			: false

enum sex { male , female }

type sex1 = (typeof sex)[ keyof typeof sex ]
type sex2 = sex.male | sex.female

type xxx = Equals< sex1 , sex > // false :(
type yyy = Equals< sex2 , sex > // false :(
type zzz = Equals< sex1 , sex2 > // true :)

🙁 Actual behavior

false for xxx and yyy.

🙂 Expected behavior

true everywhere.

[MartinJohns]

Playground link

[ahejlsberg]

Had a chance to take a quick look at this one. The example cleverly exploits one of the few places we use isTypeIdenticalTo in the compiler--which so happens to be part of a check for conditional type compatibility where we require the extends types to be identical. As the example reveals, we don't consider a union enum type (such as sex in the example) to be identical to a manually constructed union of its constituents. I'm pretty sure no one really cares about this distinction, so I'm inclined to say this is working as intended.

[ahejlsberg]

That said, I see now that this is a regression. Makes me more inclined to think we should fix it.

[nin-jin]

I'm using it in tests of parse wild json to more precise type.
Here I take all possible values of enum.
And here it fails type test.