Rethinking Rest Destructuring on Generic Targets

#62572

function f1<T extends object>(obj: T, key: keyof T) {
    const { [key]: removedValue, ...rest } = obj; // currently, `rest` is `Omit<T, keyof T>`
    return rest;
}

• rest is Omit<T, keyof T>, which is effectively object (or worse/weirder because it's like Record<never, ...>).
• Would want rest to be Partial<T> instead of Omit<T, keyof T> so that it is still useful.
• What patterns does this generalize to beyond something as specific as using T and keyof T?
• Isn't the right thing to add a K extends keyof T instead?
  • Why didn't the original code do that?
  • Probably just didn't think to add the type parameter?
    • Could imagine that knowing this specific pattern would lead to us conjuring up a type parameter bounded to keyof T so that we could get Omit<T, (some specific keyof T)>.
• We have a more precise approach, it's not the worst that we don't automatically model this specific code.
• Aside, why don't we have homomorphic Omit?
  • Design Meeting Notes, 3/8/2022 #53169

Change default --target to esnext esCURRENT

#62198

• Double-checking: are we okay with this idea?
• Breaks people without specified target.
• Kind of weird because there is stuff that is "speculative" because it's only stage 3.
• Would be better if there was eslatest instead of esnext that was more stable.
  • Maybe esstable, escurrent, esmodern, esstandard?
• Basically just refers to the latest stable version of ECMAScript.
• Probably would want the same for lib?
• So the idea is that eslatest is just an alias for e.g. es202X.
• We feel like having either esnext or some eslatest be the default makes sense because anything other than "oldest" is arbitrary.
• Maybe we don't need a specific name here, but we just default to the latest stable version of ECMAScript.
  • So this year it's es2025, next year it's es2026, etc.
• This helps with the concern about code that uses modern features, but expects syntax that most runtimes don't support to be downleveled.
• By the way, do we need to change --init?
  • Probably - either omit the target or set it to the latest stable version.