Make Map constructor argument optional and nullable

[Pimm]

Fixes #37779.

This PR is based on #38153 by @JaredNeil.

As demonstrated by the change in the baseline file, this affects the typing of new Map<string>. I don't quite understand why that is the case. I also don't know whether this is an issue, as new Map<string> (note the sole type argument) does not compile anyway. If this is no good, please point me in the right direction.

[ghost]

All CLA requirements met.

[Pimm]

With this patch applied and strictNullChecks on:

const potentiallyUndefinedIterable = [['1', 1], ['2', 2]] as Iterable<[string, number]> | undefined;
new Map(potentiallyUndefinedIterable); // Map<string, number> ✔️

new Map(undefined); // Map<unknown, unknown> ❓

new Map(null); // Map<unknown, unknown> ❓

This behaviour ‒ where undefined and null lead to Map<unknown, unknown> seems consistent with non-strictNullCheck behaviour. However, perhaps Map<any, any> is preferred. What are your thoughts?

[orta]

Yeah, seems reasonable to me, this implementation also matches how we do it in lib.es2015.collection.d.ts

interface MapConstructor {
    new(): Map<any, any>;
    new<K, V>(entries?: readonly (readonly [K, V])[] | null): Map<K, V>;
    readonly prototype: Map<any, any>;
}
declare var Map: MapConstructor;

[sandersn]

@orta this looks ready to go. Were you holding it for 4.4?

[genshinw]

I have the same issue, Is this problem has resolve?

[Pimm]

Until this lands in TypeScript, you can use ?? [] as a workaround. Like so:

const potentiallyUndefinedIterable = [['1', 1], ['2', 2]] as Iterable<[string, number]> | undefined;
new Map(potentiallyUndefinedIterable ?? []);

[orta]

I think this is good to go 👍🏻