JSDoc type ...*? is treated incorrectly

[sandersn]

I think that ...*? is equivalent to the typescript type any[] | null. However, it results in the type any[].

// @checkJs: true
// @allowJs: true
// @strict: true
/**
 * @return {...*?}
 */
function f() {
    return null // fine, but should be an error? maybe?
}

Expected behavior:

I'm not actually sure what the correct behaviour is:

1. Parse error
2. any[] | null
3. (any | null)[] (but this would mean that the prefix type parsed after the suffix type, which is incorrect).

And I'm not sure whether the error is in the parser or in getTypeOfTypeNode.

Also note that ...*= doesn't parse at all, but this is less bad (I think?).

[DanielRosenwasser]

@dimvar any tips on what Closure does here?

[dimvar]

We treat it as a parse error. We only handle ... for variable arguments, not elsewhere (wiki)

[DanielRosenwasser]

Thanks! Also sounds like you don't have syntax like A[] which means you also don't have to handle ambiguities like ?A[], correct? We might have to figure it out, either from usage or decide it ourselves.

[dimvar]

Right, we don't handle the array syntax either.

[sandersn]

I'm going to close this because

1. I couldn't find any uses of ...*? in our user test suite, although there were tons of ...*, which we handle correctly.
2. Nobody else has requested it.
3. ... is only allowed on params now anyway.
4. ...*= is now allowed, and behaves the same as ...*? -- which is the same as ...*.
5. There's not much harm in not giving an error on the extra ?/=.