optimization: refactor LastArrayElement to avoid recursion

[ahrjarrett]

Hello type-fest!

I was poking around the repo and noticed that LastArrayElement was using recursion to iterate until reaching the final element, which is unnecessary since TS 4.2:

Playground

Since people are paying more attention to type-level performance lately, this change seemed like a quick and reasonable PR to make.

[ootidea]

These are my opinions.

LastArrayElement<[]> should be undefined, but it is never.
LastArrayElement<T[]> should be T | undefined, but it is T.
LastArrayElement<[...1[], 2]> should be 2, but it is 1 | 2.
LastArrayElement<[1, 2?]> should be 1 | 2, but it is never.
LastArrayElement<[1?, 2?]> should be 1 | 2 | undefined, but it is never.
LastArrayElement<[0, 1, 2?, ...3[]]> should be 1 | 2 | 3, but it is 0 | 1 | 2 | 3 | undefined.
LastArrayElement<any> should be any, but it is unknown.

related: #635, #636

[ahrjarrett]

Personally I think those should be handled in a separate PR. The purpose of this PR was solely to make the type not recursive.

Unless, do any of these represent regressions from LastArrayElement's current behavior @ootidea ?

[ahrjarrett]

Also while I agree with some of your opinions, others seem useful only in certain use cases.

For instance, the expectation that LastArrayElement<[]> return undefined is useful only as the return type of a term-level JavaScript function that gets the last element of an array, or returns undefined if the array is empty.

My 2 issues with this:

1. Array.prototype.last does not exist (AFAIK). You could write a function called last that behaves the way you describe -- but you could also write a function that returns an error message if the array is empty, is throws, or does any number of things instead. Choosing undefined out of any other option feels somewhat arbitrary.

2. This doesn't seem to be in alignment with the design goals of type-fest, since it doesn't export (AFAIK) any term-level API. If someone tried to compose LastArrayElement with another type-level function, getting undefined would be very surprising behavior, because the only way that would happen would be if the value undefined was in the array.

[sindresorhus]

I agree with @ahrjarrett. That's out of scope for this pull request.

[sindresorhus]

@ahrjarrett Would be great if you could copy-paste your comments to the relevant issues instead: #635, #636