Type inference: is that a bug?

[jcayzac]

Acknowledgement

• I acknowledge that issues using this template may be closed without further explanation at the maintainer's discretion.

Comment

I'm not sure if the following is a bug or if it's working as intended.

interface Entry<P, D> {
  point: P
  datum: D
}

class X<P, D, E extends Entry<P, D>> {
  readonly p: P
  readonly d: D

  constructor(entry: E) {
    this.p = entry.point
    this.d = entry.datum
  }
}

const a = new X({ point: 1, datum: 'a'})

Here, a is inferred to be of type X<unknown, unknown, { point: number; datum: string; }> (while a.p and a.d have the correct types).

I expected it to be of type X<number, string, { point: number; datum: string; }>.

Can anybody explain what's going on here? I'm using Typescript 5.4.3.

[Andarist]

This is just not something that TS supports. Inference targets have to appear in parameter lists - the "sources" in constraints of other type parameters are ignored. An even simpler repro:

interface Box<T> {
  value: T;
}
function test<T, B extends Box<T>>(a: B) {}
test({ value: 10 });

Arguably, you probably don't want this signature. In situations like this you probably don't want to allow such calls:

new X<number, string, { point: 1, datum: 'a' }>({ point: 1, datum: 'a' })

I'd assume that your E should be strictly { point: number, datum: string } here but your signature allows different subtypes of that.

[jcalz]

see #7234 which was not implemented in favor of having people use intersections instead.

[jcayzac]

Thank you both, @Andarist @jcalz