Type errors on nullable in Typescript 4.9 about pothos

[estyrke]

I just verified - the error does not appear in TS 4.8.4, and it is still there in 4.9.5.

[estyrke]

If I change the nullable to nullable: { items: true, list: false }, the error becomes (the possibly more clarifying)

error TS2322: Type '{ items: true; list: false; }' is not assignable to type '({ items: true; list: false; } & true) | ({ items: true; list: false; } & { items: boolean; list: boolean; } & true) | undefined'.
  Type '{ items: true; list: false; }' is not assignable to type '{ items: true; list: false; } & { items: boolean; list: boolean; } & true'.
    Type '{ items: true; list: false; }' is not assignable to type 'true'

[hayes]

I tried to reproduce this issue, but have not been able to:

https://stackblitz.com/edit/typescript-rmf22w?file=index.ts,tsconfig.json

This sees specific to types not included in the example above

[estyrke]

Thanks for looking into it! After trying for quite a while to extend the stackblitz enough without succeeding, I finally realized that explicitly typing the return value of Group.children to string[] resolves the issue.

I'm a bit curious why the stackblitz differs from my code though. I tried

export class Group {
  get children(): ("a" | "b")[] {
    return ["a"]; //GROUP_DEFINITIONS[<GroupName>this.name];
  }
}

type Foo = ExposeNullableOption<
  PothosSchemaTypes.ExtendDefaultTypes<{}>,
  ["String"],
  Group,
  "children"
>;

In the "real" code (using VS code tooltips), Foo resolves to

type Foo = FieldNullability<["String"]> & (true | {
    items: boolean;
    list: true;
} | {
    items: boolean;
    list: true;
})

But in stackblitz it's

type Foo = FieldNullability<["String"]> & (boolean | {
    items: boolean;
    list: boolean;
} | {
    items: boolean;
    list: boolean;
})

I'm no typescript expert, but It seems like this could be a typescript bug.

type Bar = Group["children"] extends NonNullable<Group["children"]>
  ? true
  : false;

resolves to true in both my local editor and in stackblitz, which I think means that nullable should be allowed to be false.

[hayes]

Thanks for the detailed reply! Figured out the issue, and fixed it here: 3021b43

[estyrke]

Works like a charm! Thanks for persisting!