Allow negated types + F-bounded mapped types

Question

Allow negated types + F-bounded mapped types

dead-claudia opened this issue 7 years ago · 5 comments

dead-claudia commented 7 years ago

Edit: keyof works as I thought; mapped types are just broken in interfaces. I also clarified a few things.

Allowing negated types and F-bounded mapped type variables would solve several issues:

Subtraction types: #4183 - The proposed A - B could be encoded as A & !B as per this proposal.
Mapped conditional types: #12424 - For key-based conditions, this + F-bounded mapped types would allow that to be solved without explicit support.
Promises - Having negated types makes it possible to prevent thenables from being used as generic Promise parameters.

Here's my proposal:

Allow negated types, using !T to represent the set of all types that do not conform to T (i.e. not assignable to T). This is pretty simple to digest.
Expand mapped types to allow the key variable to be F-bounded, like in {[P in Keys<T, P>]: ...}.

The first is sufficient to cover subtraction types and Promise type issues, and the other two make mapped conditional types possible.

Answer 1 · 2017-09-06T17:55:54.000Z

Big 👍 on the proposed syntax. I think the negation operator is indeed quite intuitive in the context of mapped types/keyof and is a great idea. My only concern is that I'm not sure what it means outside of an intersection context. "Does not conform to T" could be construed as meaning not assignable to T which is very different from meaning not having any of the keys of T. This may just need further clarification.

Minor nitpick, but I believe your second item is already valid today.

Answer 2 · 2017-09-07T01:48:30.000Z

@aluanhaddad

My only concern is that I'm not sure what it means outside of an intersection context.

This comment should explain type negation outside a union context (partial Promise fix).

"Does not conform to T" could be construed as meaning not assignable to T which is very different from meaning not having any of the keys of T.

"not assignable to T" is actually the intended meaning for negated types. This happens to be equivalent to "not having any of the keys of T" in the context of object types, but it carries more expected semantics for primitives.

Minor nitpick, but I believe your second item is already valid today.

Yeah...I figured this out; I just ran into this odd bug.

Answer 3 · 2018-03-09T04:52:35.000Z

I upvote A & !B especially the !B part... Over Exclude from #21847

Answer 4 · 2018-03-09T17:55:48.000Z

Looks like the same feature as #4196, just different proposal. i suggests adding the proposal to #4196 and keeping the discussion to one issue.

Answer 5 · 2018-03-23T19:03:33.000Z

Automatically closing this issue for housekeeping purposes. The issue labels indicate that it is unactionable at the moment or has already been addressed.