Variadic tuple constraint incoherent ?

[lifaon74]

TypeScript Version: 4.0.0-dev.20200722

Search Terms:
Variadic tuple
Variadic tuple constraint

Code

type A<GArgs extends [number, string, ...any[]]> = true;
type B<GArgs extends [string, ...any[]]> = A<[number, ...GArgs]>; // ERROR
type C = A<[number, ...[string, ...any[]]]>; // OK

Actual / Expected behavior:

On the type B we get an error:

TS2344: Type '[number, ...GArgs]' does not satisfy the constraint '[number, string, ...any[]]'.   Element at index 1 is variadic in one type but not in the other

Sounds incorrect because on B, GArgs is properly shaped to be just like in C

Playground Link

Related Issues:

The variadic issues are a big jungle, and I was not able to find any related issue (but probably some exists?)

[ahejlsberg]

Our relationship reasoning for variadic tuples is currently quite conservative, and we only match variadic elements against other variadic elements. In the example above, it is indeed the case that any instantiation of [number, ...GArgs] would satisfy the constraint [number, string, ...any[]], as is evident by substituting the constraint [string, ...any[]] in place of GArgs. However, it isn't generally the case that a variadic element's constraint can be substituted in its place. For example:

type A1<T extends [number, string?, boolean?]> = true;
type B1<U extends [string?]> = A1<[number, ...U, boolean?]>;

Here, the instantiation B1<[]> corresponds to A1<[number, boolean?]> which doesn't satisfy the constraint.

Certainly, in cases where a variadic element occurs in the last position of a tuple, and cases where a variadic element has a fixed length tuple constraint, we could do additional reasoning. So, I'll label this issue a suggestion to that effect.