Mapped types fixes

[gcnew]

Fixes #17238

This fix changes the rule from #12351 to:

• The operation { [P in K]: T}[X] is equivalent to an instantiation of T where X is substituted for every occurrence of P, unless K is an indexing operation itself. For example, { [P in K]: Box<T[P]> }[X] is equivalent to Box<T[X]>, however { [P in T[X]]: Box<T[P]> }[X] is not.

This PR also fixes checking of operations done on the object of type indexing, e.g. the following is now correctly identified as error:

// Note that `extra` is not present in { a:'a', b:'a' }
type T5<S extends 'a'|'b'|'extra'> = {[key in { a:'a', b:'a' }[S]]: true}[S]

Examples:

type AB = {
    a: 'a'
    b: 'a'
}

type T1<K extends keyof AB> = { [key in AB[K]]: true }
type T2<K extends 'a'|'b'> = T1<K>[K] // BUG 1: should be error for K = 'b'

type R = AB[keyof AB]; // "a"
type T3 = { [key in R]: true }
type T4<K extends 'a'|'b'> = T3[K] // error as expected

// BUG 2: 'extra' not checked in AB[S]
type T5<S extends 'a'|'b'|'extra'> = {[key in AB[S]]: true}[S]

[DanielRosenwasser]

@ahejlsberg can you take a look at this PR?

[ahejlsberg]

@gcnew Thanks for your contribution! You are definitely right about the missing checkSourceElement call call in checkIndexedAccessType. In fact, there should be a call for node.objectType as well.

Regarding the other change, the correct fix is to add a check in getIndexedAccessForMappedType that verifies that the index type is assignable to keyof T for the object type. This is supposed to be true for any indexed access type but ends up not being verified in the generic mapped type case.

I'm going to put up a new PR with these changes, but I will keep the tests you wrote. Thanks again!

[gcnew]

Funny enough, I tried to fix checkIndexedAccessType at first, but due to my lack of knowledge and guided by the debugger, I went for getIndexedAccessType.

Now, I'm looking at another issue. Mapped types buried inside an intersection are eagerly resolved, leading to incorrect behaviour (see #17456 and gcnew@73c78e6). The problem is that if the indexing operation is postponed (as in my experiment), the index type is not properly checked.

[gcnew]

Closing as the bugs have already been addressed by #17455.