Number.is* should accept unknown

[shicks]

TypeScript Version: 3.8.0-dev.20191102

Search Terms: Number isInteger isSafeInteger isFinite isNaN

Code

declare function use(number: number): void;
declare const maybeNumber: string | number;
if (Number.isInteger(maybeNumber)) {
  use(maybeNumber as number);
}

Expected behavior:
Type checks

Actual behavior:
Fails to type check because Number.isInteger is typed to take only number.

Playground Link:
http://www.typescriptlang.org/play/index.html?ts=3.8.0-dev.20191102&ssl=3&ssc=2&pln=1&pc=1#code/CYUwxgNghgTiAEAzArgOzAFwJYHtX2QGcQAKVZAWwCMQYAuec62gSgYDcctgBuAKFCRYCMHkIZ4FKAE8aAOUo168cTCyoA5vAA+jRbX5ZE8EguYwAdFkIBJVBhAbaJKbJBmlLFvADefeATELjLy+jDwUIR65iz8AL5AA

Related Issues:

• Fixes Number.is* signatures to accept any input. #24436 - PR that already addressed this 18 months ago, but seems to have been accidentally reverted by 2f73986.
• lib.d.ts contains wrong typings for Number.isFinite() #15527 - closed for housekeeping.
• isNaN and isFinite should accept number or string #4002 - not the same issue - that issue is about the global functions that coerce their argument, while this is about the functions on Number that don't, making it safe to widen the accepted types.
• Number.isInteger and others should accept any and act as type guards #21199 - a broader issue, since it's currently infeasible to make these type guards.

[shicks]

Reading further, it looks like #24436 updated the generated lib, not the version in src.

[RyanCavanaugh]

Thank you for the good research effort on this!

[DanielRosenwasser]

So do we want this change? I'm kind of glad it didn't happen (even for bad reasons) because I think the current behavior saves users more than it frustrates, and the workaround is relatively simple (unary +).

[RyanCavanaugh]

@DanielRosenwasser I think it's good. Number.isInteger("0") is false, IOW no fishy coercion occurs