Support Array.prototype.at

[bpasero]

There is a new method being added for Array: at

https://developer.mozilla.org/en-US/docs/Web/JavaScript/Reference/Global_Objects/Array/at

I think typing this in a way that undefined is returned would allow a gradual adoption of the noUncheckedIndexedAccess rule:

at(index: number): T | undefined;

[MartinJohns]

I think typing this in a way that undefined is returned would allow a gradual adoption of the noUncheckedIndexedAccess rule:

I don't see how this is related to noUncheckedIndexedAccess. The function returns an explicit undefined if the index does not exist.

Also, please note that there is an issue template for library changes: lib_change.md

[bpasero]

I don't see how this is related to noUncheckedIndexedAccess.

Being in a codebase that would produce hundreds of compile errors when noUncheckedIndexedAccess is enabled, adopting Array.at would allow to gradually adopt this to the point where noUncheckedIndexedAccess could then be enabled by default.

I notice #36554 exists, should I merge my issue into there?

[SimonSchick]

You can always just add ! null assertions to your array indexing, and gradually remove it, at shouldn't be used as a migration strategy imo.

[bpasero]

👎 For me, using ! is like using any and I stay away from it unless there is really no other way.

[FelipeSharkao]

The biggest thing of Array.prototype.at is the possibility to use negative indexes. array.at(-1) is way clearer than array[array.length - 1].

[orta]

We chatted in the design meeting about this #45551, and agreed with @bpasero that it's a good call (and we debated techniques for typing the minus numbers but don't have an instant answer yet)

Also linking to #44656 which this should probably be a part of

[Ethan-G]

For impacts: I went to use .at() today only to discover it's not supported yet.

[jonlepage]

ho i lost so many time to try understand !

So It not yet added ! ?

[MartinJohns]

Also linking to #44656 which this should probably be a part of

@orta You probably mean #44571.

[unicornware]

@Ethan-G not sure if you found a workaround yet, but i ended up adding a custom .d.ts file (i also tried upgrading to typescript@4.5.0-beta, but found that Array.prototype.at wasn't supported in that version either):

// typings/lib.esnext.array.d.ts

/** @template T - Array item type */
interface Array<T> {
  /**
   * Takes an integer value and returns the item at that index, allowing for
   * positive and negative integers. Negative integers count back from the last
   * item in the array.
   *
   * @param {number} index - Position of the array element to be returned
   * @return {T | undefined} Element in the array matching the given index.
   * Returns `undefined` if the given index can not be found
   */
  at(index: number): T | undefined
}

/** @template T - Array item type */
interface ReadonlyArray<T> {
  at: Array<T>['at']
}

docs pulled from MDN.

[PabloLION]

I'm with typescript@4.5.3 yet still having the same problem as @djmisterjon did.
In the playground it shows the same. How should I configure TS to suppress the error?

[PabloLION]

It seems I would have to wait until es2022 lib comes out?

[jonlepage]

it added in 2022 lib

[PabloLION]

I can't get ES2022 lib in my TSconfig with 4.5.3.. I'll wait for 4.6

[jonlepage]

I can't get ES2022 lib in my TSconfig with 4.5.3.. I'll wait for 4.6

ho you may true, i using 4.6-dev in my workspace

[hochan222]

Thanks to those who raised the issue in the first place!

[PabloLION]

As for now, this functionality is supported by typescript@4.5.4, while setting the tsconfig.module to es2022.

[RebeccaStevens]

Would it be possible to change the return type of at for tuples?
For example make it so that ([1, 2, 3] as const).at(1) will be 2 rather than number | undefined?

Or would this be too difficult considering that we'd need to take into account how negative indexes work?

[jonlepage]

hum yes , maybe they will need change the way how the generic share value.
Maybe try open a new post
Here a hack solution where is work !
edit: not handle negative index

interface ReadonlyArray<T> {
    /**
     * Returns the item located at the specified index.
     * @param index The zero-based index of the desired code unit. A negative index will count back from the last item.
     */
    at<U extends number>(index: U): this[U] | undefined;
}

[PabloLION]

I agree with @djmisterjon 's idea on new post.
But here's the code I used. const a = (([1, 2, 3] as const).at(1 as const))!
BTW @djmisterjon I find your "hack solucion" doesn't work. maybe you used a T and a U?

[PabloLION]

The ! I used was for the at() method cannot check if the element exists, so a | undefined is always deduced. I remembered that this problem is reported.

[jonlepage]

i dont know what to say, worked on my side with the hack

And if you remove as const it become number

So for me its should natively support this without manual hack.

[PabloLION]

I think it's working correctly. TS deduces that the type of [1,2,3] is number[], like [] as never[]. What do you suggest here..?

[MartinJohns]

@djmisterjon That's not a hack, that's the expected and desired behaviour.

[jonlepage]

...heum ,hi, i think you don't understand the meaning of the word hack in this context.
Hack mean bidouillage and not hacking.
Hack your code or hack this code can mean add, edit or modify
Modify (hack) the native code provided by default to test a new behavior.