Type inference issue with spread and generics

[flovilmart]

We notice the following issue described in this playground.

TL;DR;

The type inferred when merging two objects trough a generic function is always resolved to T & U which is inexact if T shares keys with U.

The correct inferred type of const mergeObjects = <A, B>(a: A, b: B) => ({ ...a, ...b });

should be

type Merge<T, U> = { 
    [k in Exclude<keyof T, keyof U>]: T[k];
} & {
    [l in Exclude<keyof U, keyof T>]: U[l];
} & {
    [l in keyof (T | U)]: T[l] | U[l] 
};
<A, B>(a: A, b: B) => Merge<A, B>

As shown in the playground, the correct type is inferred when using the spread operator with concrete types.

[RyanCavanaugh]

The proposed definition is much more complex yet still not fully accurate because it can't account for aliasing effects.

See discussion in #28234

[typescript-bot]

This issue has been marked as a 'Duplicate' and has seen no recent activity. It has been automatically closed for house-keeping purposes.

[flovilmart]

Thanks for pointing it out.