I"d expect this to compile:

type T<V>     = typeof valueProducer<V>  // TS2304: Cannot find name "V".
// or
type T<V>     = (typeof valueProducer)<V>

and I"d expect T<V> to be exact as ValueProducer<V>

Or, I"d expect:

type T = typeof valueProducer  // Error - Type `T` is not generic
type T<A> = typeof valueProducer  // Compiles fine, `A` is bound to `V` in `valueProducer`

The two types are actually different. The way you have declared valueProducer, you are saying it is a function which takes in a parameter of any type and returns the same type as the parameter. (This is what <V>(i?: V) => V means). However, ValueProducer<V> is a function which takes in a parameter of one specific type (V) and returns V. Just as ValueProducer is a meaningless type without the generic argument, typeof valueProducer is not a generic type and so it is meaningless to give it a generic argument.

@calebsander Exactly, I propose to actually make the type-level typeof to return the "real", "open" generic types in such cases.

Otherwise there"s an inconsistency in the language. You are saying that valueProducer

is a function which takes in a parameter of any type and returns the same type as the parameter.

In other words, its a "normal" type, not a generic. Yet it is possible to provide a type argument for the valueProducer during the call: valueProducer<Date>(new Date())

Why is it so, if the type of valueProducer does not have any type arguments? Special case? Exception from rules? Why, if its possible to just make typeof to be aware of generic arguments.

I think this is related: #17574

@jack-williams it is, yes, though the scope of the latter is broader than this ticket. Using the terminology from #17574, it can be formulated as:

function id<V> (v : V) { return v } 

type Id2<V>  = (v: V) => V
type Id3     = <V>(v: V) => V

type TypeOfId = typeof id

Which one of the Id2 and Id3 should be returned by typeof id? Currently it is Id3 and I think the correct answer is Id2, because you can provide a type argument for the id during the call.

Which one of the Id2 and Id3 should be returned by typeof id? Currently it is Id3 and I think the correct answer is Id2, because you can provide a type argument for the id during the call.

The type of id should be Id3 because values should be represented by types like <V>(v: V) => V, not functions from types to types.

Id2 and Id3 really are not the same. The first one just introduces a type name for sharing in the type graph, the second denotes something that is genuinely parametric in type. Just because I can write:

type Id<V>  = (v: V) => V
const f: Id<boolean> = (x: boolean) => !x; // not parametric

this does not mean that f is, or ever was, parametric in the type of its argument. However the following f is, hence why I can instantiate it freely.

type Id<V>  = (v: V) => V
const f: Id<boolean> = <X>(x: X) => x; // parametric

Following your suggestion to return Id2 would actually be weakening the type because you end up "forgetting" that the value was ever actually parametric. For instance, I could not do this.

type NumberPicker = (picker: <X>(x: X, y: X) => X, x: number) => number

const constantPicker1 = <X>(x: X, y: X) => x;
const constantPicker2 = <X>(x: X, y: X) => y;

declare const x: typeof constantPicker1;
declare const y: typeof constantPicker2;

const pick: NumberPicker = (picker, x) => picker(x, -x);

// ok as x and y are parametric
pick(x, 3);
pick(y, 4);

I think what you really want is to separate type application from function application, so it can be done partially, and then have that syntax in type queries. Like:

type T<V> = (typeof valueProducer<V>)

I think what you really want is to separate type application from function application, so it can be done partially, and then have that syntax in type queries. Like:

type T<V> = (typeof valueProducer<V>)

This will work for my purposes, yes! Any chance for this to be implemented?

This issue has been marked "Working as Intended" and has seen no recent activity. It has been automatically closed for house-keeping purposes.

@jack-williams FYI, I"ve filed the feature request for solution you suggested: #29043

It seems to be small, non-breaking and very useful feature, hope to see it implemented.