Inconsistent behaviour of -?

[m-shaka]

🔎 Search Terms

optional mapped type

🕗 Version & Regression Information

• This changed between versions 5.4.5 and 5.5.4

⏯ Playground Link

https://www.typescriptlang.org/play/?exactOptionalPropertyTypes=true#code/C4TwDgpgBA8mwEsD2A7AhgGwCpIKouRQHUFgALfAEwgDMEUJKAeLAPigF4oBvAKCgFQA2gGko9KAGsIIJDShYAugC4FUCAA9gEFJQDOUAEoQAxkgBOzEQBoFoxa36DnAfjsjFT5wNVZ7UAB8oAFddWnpGXgBfXl5QSCgASRRtcwBbBEoENG1OWHhCTBx8QhJyKnCGZm4oNBdVPWBzegBzKCj2AHpOnlr6qEbmlDag0Oo6Kva48GgAGVIIc0w8mrqGptbAkLCJxnaoboG0NOh46DQDZNSMrJyIWLOjCD0ARjzjAEdghHNGJivFjdstouj1Vushm0Yo9jHoAEzvCBfH5-eapTCg3poCGbUY7CKUdpAA

💻 Code

type OptionalToUnionWithUndefined<T> = {
    [K in keyof T]: T extends Record<K, T[K]>
        ? T[K]
        : T[K] | undefined
}

type Intermediate = OptionalToUnionWithUndefined<{ a?: string }> // { a?: string | undefined }
type Literal = { a?: string | undefined } // same type as Intermediate

type Res1 = Required<Intermediate> // { a: string }
type Res2 = Required<Literal> // { a: string | undefined }

🙁 Actual behavior

Res1 and Res2 are different types

🙂 Expected behavior

they should be the same because the inputs are

Additional information about the issue

This issue may share the cause with #59902

[Fullfungo]

I think TS is tripping balls and hallucinating here. Both Res1 and Res2 are supposed to be {a: string | undefined}, and they are!
The following code compiles:

type Res1a = Res1['a']; // string | undefined
let r: Res1 = {a: undefined}; // Ok

type Id<T extends object> = {[K in (keyof T & string)]: T[K]};
type Res1Id = Id<Res1>; // { a: string | undefined }

This issue might be related to the --exactOptionalPropertyTypes flag. However, Required works differently without the flag, so I am not certain.

[RyanCavanaugh]

There's no -? in the code sample; what's the relation to the title of the issue?

[Fullfungo]

There's no -? in the code sample; what's the relation to the title of the issue?

The -? is in Required. Hope this helps.

[RyanCavanaugh]

I think this is just a type display issue. If you actually access the property, it has the expected type:

type OptionalToUnionWithUndefined<T> = {
    [K in keyof T]: T extends Record<K, T[K]>
        ? T[K]
        : T[K] | undefined
}

type Intermidiate = OptionalToUnionWithUndefined<{ a?: string }> // { a?: string | undefined }
type M = { a?: string }
type Literal = { a?: string | undefined } // same type as Intermidiate

type Res1 = Required<Intermidiate> // { a: string }
type Res2 = Required<Literal> // { a: string | undefined }

declare let r: Res1;
const j = r.a;
// j: string | undefined

// OK
const p: Res2['a'] = undefined;

[mkantor]

@RyanCavanaugh In TypeScript 5.5.4 (the regressed version mentioned in the issue) and newer your example has a type error, which indicates this is more than just a display issue. EDIT: Wait, I may have gotten confused by the comments in your example which don't match the type info I'm seeing. Res1 and Res2 appear to be the same in both type info and behavior in the playground above when using TypeScript 5.5.4. Is that expected?

Here's a reduced reproduction (I think).

[m-shaka]

@mkantor Your first example is related to exactOptionalPropertyTypes option

https://www.typescriptlang.org/play/?exactOptionalPropertyTypes=false&ts=5.5.4#code/PTAEG9QQwLlBnALgJwJYDsDmoA+oCu6AJgKYBmGJRoAvqKmaCQB5QDGiA8gA6KoD26KABsACsn7cSyRAE8AKrKnwAvCnwkAUCAjQ4SNFlr1GLdl14ChYiVJkKlJVWRHwtcqaHmgVoAEokAI74qMhUADyQUAD8+igY2HiEpBToVLQAfJqabIJIel4A2gDkUMUAuj4ExOSUREA

For my example, which enables exactOptionalPropertyTypes , Res1 should be displayed { a: string | undefined } but not

[mkantor]

Your first example is related to exactOptionalPropertyTypes option

Ah, right you are! I copied the prior example into a playground without enabling that option. Thanks for catching my mistake.

I believe my latter reduced example is the same bug though (despite not requiring exactOptionalPropertyTypes). I verified that the proposed fix at #59957 addresses it. I think it's essentially the same scenario that #60411 is talking about.

[m-shaka]

I see, the latter one looks like truly minimum reproduction of this bug! Thank you