Inconsistency with Rest Parameter in Function Type

[nlipiarski]

🔎 Search Terms

rest parameters, parameter list, function parameters

🕗 Version & Regression Information

• This is the behavior in every version I tried, and I reviewed the FAQ for entries about parameters and rest parameters

⏯ Playground Link

https://www.typescriptlang.org/play/?ts=5.4.5&ssl=5&ssc=12&pln=1&pc=1#code/C4TwDgpgBAysBOBLAdgcwGIFdkGNiIHtkoBeKACgDpqBDeVAZwC4oGEVUBtAXQEpSAfFABuBRABMA3AChpOImygQAHjQC2YADYQAjCzhI0WXPiKkKdVHtbs0-EkPnIGBbZU0FU5SzvcQ0wAAWvDLSKupauuQhQA

💻 Code

type StringFunction = (...args: string[]) => void;

const example1: StringFunction = (arg1: string) => console.log(arg1.length);

example1();

🙁 Actual behavior

This code does not display any type errors on any of the versions I tried, but will break at run time on the example1() call.

🙂 Expected behavior

Typescript should prevent the assignment of (arg1: string) => void to a variable of type (...args: string[]) => void since it is possible to have args be empty.

Additional information about the issue

I attempted to search for other issues related to this and came up empty handed. This is different from the "I'm allowed to use a shorter parameter list where a longer one is expected" issue which has drowned out a lot of my searching.

Thanks in advance for any help with this.

[Andarist]

I found this in the existing tests:

TypeScript/tests/cases/conformance/types/typeRelationships/assignmentCompatibility/assignmentCompatWithCallSignaturesWithRestParameters.ts

Line 16 in 79a8514

a = (x: number) => 1; // ok, rest param corresponds to infinite number of params

This is a pretty old test though and it assumes strictNullChecks: false. I think it doesn't quite make sense in the strict world for this to be allowed. Rest params are effectively optional so why do we observe a difference here?

// error
const fn1: (a?: string) => void = (a: string) => {}

// ok?
const fn2: (...args: number[]) => void = (x: number) => {};

We can find some recent tests with strict: true that would break if the above wouldn't work:

TypeScript/tests/cases/compiler/assignmentToAnyArrayRestParameters.ts

Lines 6 to 14 in 79a8514

	function foo<T extends string[]>(
	fa: (s: string, ...args: string[]) => string,
	fb: (s: string, ...args: T) => string
	) {
	const f1: (...args: any) => string = fa;
	const f2: (...args: any[]) => string = fa;
	const f3: (...args: any) => string = fb;
	const f4: (...args: any[]) => string = fb;
	}

At the very least f2 and f4 should error here - assuming that the issue at hand would get fixed.

[fatcerberus]

rest param corresponds to infinite number of params

Yeah, no. It can be up to an infinite number of params, but a minimum of zero. i.e. the same reason fn1 is an error in your example.

I guess the reasoning here is the same as how an object with a string index signature is treated in some contexts as having all possible properties at once, which is only rectified by enabling noUncheckedIndexedAccess.