Expose ArrayLength type

[jericirenej]

Issue #1341 proposed either deleting or exporting the unused ArrayLength internal type.

The issue was marked as a duplicate of #676 with the comment that exposing internal types is encouraged.

In this PR:

• ArrayLength is moved from ìnternal/array.d.ts to a separate file in source.
• General tests added.
• Type documentation has been expanded.

[som-sm]

Useful for enforcing fixed-length arrays

What do you mean by this? Can you also add an example for this.

distinguishing between tuple and non-tuple like arrays

I suggest removing this because we already have the IsTuple type, which handles this case better. For example, IsTuple allows [number, ...string[]] to be treated as both a tuple and a non-tuple, whereas ArrayLength can only treat it as a non-tuple.

[jericirenej]

Yes, it's a bit vague, I agree!
For enforcing fixed-length array this meant that it could have been used as part of a combined type that would be checking against some other numeric literal condition. For example:

type IsLength<
  Arr extends readonly unknown[],
  Num extends number,
> = IsEqual<ArrayLength<Arr>, Num>;

But this might all be a bit too contrived. And I agree with your point about the tuple / non-tuple distinction - the IsTuple helper does a much better job.

I'm fine with removing the "usefulness" sentence if it is not needed. I added it to replicate the documentation pattern observed in some other helper types. But the simplicity of the current one might make it hard to do so.

p.s.: I was resolving the conversations /comments as I was applying the changes - I should have left it to you to resolve it, since you made the actual change requests. Sorry about that, I was on auto-pilot 😄 Please feel free to unresolve any of them if I made an error somewhere.

[som-sm]

I think you can remove this entirely. Adding contrived use cases doesn't really help, and it's a pretty simple type, so it doesn't necessarily need a use case section.

[jericirenej]

Agreed. I'll remove and push

[jericirenej]

Strange, the IDE inferred unknown when specifying ArrayLength<any>, but tests failed, because there it was inferred as an any type. Updated.

[som-sm]

LGTM! Just few minor comments.

[som-sm]

I think you can remove this entirely. Adding contrived use cases doesn't really help, and it's a pretty simple type, so it doesn't necessarily need a use case section.

[sindresorhus]

https://github.com/sindresorhus/type-fest/pull/1344/changes#diff-a6653dddc77761046fd084a2c26be9d7c4af1eaf6fc2394dc42d8d7f67281e3cR34 fails.

[som-sm]

@jericirenej Updated the implementation of ArrayLength to the following:

export type ArrayLength<T extends readonly unknown[]> = T['length'];

type T = ArrayLength<any>
//=> any

The previous implementation returned unknown when instantiated with any. IMO, returning any is better.

type ArrayLength<T extends readonly unknown[]> = T extends {readonly length: infer L} ? L : never;

type T = ArrayLength<any>
//=> unknown

Also, the updated implementation is much simpler.

[jericirenej]

@som-sm Looks great, thanks! 👍
I was also curious initially why the original type was inferring the length instead of just accessing length directly. But at that point I did not venture further and just focused on providing some tests and documentation which would satisfy the requirements for the type to be exported.

This possibility of accessing the length directly was also one of the reasons that - when faced with the options to expose the type or remove it - I was leaning more into arguing for its omission in my initial post.

[som-sm]

This possibility of accessing the length directly was also one of the reasons that - when faced with the options to expose the type or remove it - I was leaning more into arguing for its omission in my initial post.

I agree, this type isn’t that useful. It’s essentially just syntactic sugar, T["length"] vs ArrayLength<T>. I’m also fine with removing it. @sindresorhus, WDYT?

[sindresorhus]

I think it's useful as documentation. It's easy to say that T["length"] is better, but most don't even know about it. I also find ArrayLength slightly more readable at callsite. So our docs can make it clar that T["length"] is also a good alternative.

[jericirenej]

Ok, I'll update the base type description and docs later today!

[jericirenej]

Rebased to current main and updated documentation.

Hopefully the added documentation is clear and concise enough:
Return the length of an array. Equivalent to `T['length']` where `T` extends any array.