Implement TypedArray.prototype.with

[robik]

Summary

This PR implements ES2023 TypedArray.prototype.with method. This is a follow-up of PR #1286 which added with method on standard Array prototype.

See also:

• PR for TypedArray.prototype.toReversed: Implement TypedArray.prototype.toReversed #1366
• PR for TypedArray.prototype.toSorted: Implement TypedArray.prototype.toSorted #1367

I am creating separate PRs for each method in case there are any suggestions or change requests from your side, so that PRs are more independent.

Test Plan

Code is annotated with algorithm steps from EcmaScript specification for easier verification and maintenance.
I also added tests to verify that methods work as intended. There might be some more edge cases that might be covered based on your experience.

$ echo "new Int8Array([0]).with(0, 5)" | ./bin/hermes
# >> Int8Array [ 5 ]

[robik]

Also, as an future improvement area, is it possible to determine if an array is an rvalue (as in a temporary value, which reference to was not captured yet)?

I was thinking of implementing a basic copy-ellision. If there's a way right now, I might create a follow-up PR that implements it, e.g: checks if the target array is an rvalue, and if so, changes it in-place for some non mutating methods so that we do not allocate new array on every call.

This would be beneficial for cases like below:

const finalArray = [0, 0, 2]
  .with(1, 1)
  .toReversed();

[facebook-github-bot]

@dannysu has imported this pull request. If you are a Meta employee, you can view this diff on Phabricator.

[tmikov]

@robik, high level optimizations like that are definitely something we are planning for in Static Hermes, but unfortunately they are almost impossible to do ahead of time in a regular untyped JS.

In regular JS, when we encounter x.with(1, 1).toReversed(), we don't know whether x is an Array, what with() does, whether it has been overridden by a polyfill, etc.

Things change dramatically in Static Hermes:

let x : any[];
...
x.with(1,1).toReversed()

[avp]

Thanks for the PR, it looks mostly good. I've left a couple comments regarding pointer safety that will have to be fixed before we can land it.

[facebook-github-bot]

@robik has updated the pull request. You must reimport the pull request before landing.

[facebook-github-bot]

@robik has updated the pull request. You must reimport the pull request before landing.

[facebook-github-bot]

@robik has updated the pull request. You must reimport the pull request before landing.

[facebook-github-bot]

@robik has updated the pull request. You must reimport the pull request before landing.