Consider the following:
type Properties = {
foo: { n: number };
bar: { s: string };
baz: { b: boolean };
};
declare function retrieveValues
There's no straightforward type operator which, say, turns a union into an intersection, or allows you to iterate union types and do stuff programmatically with the pieces. So on the face of it you're stuck.
Backing up, if you allow yourself to build Properties
from pieces instead of trying to break the pieces apart, you can do this:
type InnerProperties = {
n: number;
s: string;
b: boolean;
}
type OuterProperties = {
foo: "n";
bar: "s";
baz: "b";
}
You can see how each key in OuterProperties
is a mapping to a key in InnerProperties
. (Note that in your Properties
, each outer property had a single inner property. You aren't restricted to that, though. If you wanted, say, the "foo"
outer key to correspond to something with multiple inner properties like {n: number, r: RegExp}
then you would add r: RegExp
to InnerProperties
and put foo: "n"|"r"
in OuterProperties
.)
Now you can pick out partial properties like this:
type PickProps<P extends keyof OuterProperties = keyof OuterProperties> = {
[K in OuterProperties[P]]: InnerProperties[K];
}
So PickProps<"foo">
is {n: number}
, and PickProps<"bar">
is {s: string}
, and PickProps<"baz">
is {b: boolean}
. And notice that PickProps<"foo"|"bar">
is {n: number; s: string}
, so we have the output type of retrieveValues()
ready. We still have to define Properties
in terms of InnerProperties
and OuterProperties
, like this:
type Properties = {
[K in keyof OuterProperties]: PickProps<K>
}
And finally you can declare that function the way you want it:
declare function retrieveValues<K extends keyof Properties>(add?: K[]): PickProps<K>;
const y: { n: number } & { s: string } = retrieveValues(['foo', 'bar']);
So that works. Hope that's helpful. Good luck!
Using conditional types and type inference in conditional types it is possible to transform { n: number } | { s: string }
directly into { n: number } & { s: string }
.
type GetKeys<U> = U extends Record<infer K, any> ? K : never
type UnionToIntersection<U extends object> = {
[K in GetKeys<U>]: U extends Record<K, infer T> ? T : never
}
type Transformed = UnionToIntersection<{ a: string } | { b: number }>
// Transformed has type {a: string, b: number}
Playground Link
The reason this works is basically because conditional types are distributed over union types. From the conditional types pull request:
Conditional types in which the checked type is a naked type parameter are called distributive conditional types. Distributive conditional types are automatically distributed over union types during instantiation. For example, an instantiation of
T extends U ? X : Y
with the type argumentA | B | C
forT
is resolved as(A extends U ? X : Y) | (B extends U ? X : Y) | (C extends U ? X : Y)
.