no-unnecessary-generics reporting in a less useful place than it could?

[johnnyreilly]

Hi all!

I've been looking into this Definitely Typed pull request and picking through the issues with @JoshuaKGoldberg. We encountered an interesting case with the no-unnecessary-generics rule which we think could be better. Consider:

    <TElement extends HTMLElement = HTMLElement>(
      html: JQuery.htmlString,
      ownerDocument_attributes?: Document | JQuery.PlainObject,
      // eslint-disable-next-line @definitelytyped/no-unnecessary-generics
    ): JQuery<TElement>;

You'll note that we have an eslint-disable-next-line just before the last line; the ReturnType of the method. Without this in place the red squigglies would be under:

JQuery<TElement>
       ~~~~~~~~

This isn't the useful place to have this report. Perhaps the better place would be the first mention of TElement, like so:

    <TElement extends HTMLElement = HTMLElement>(
     ~~~~~~~~

Looking at

DefinitelyTyped-tools/packages/eslint-plugin/src/rules/no-unnecessary-generics.ts

Line 57 in 822dadb

node: parserServices.tsNodeToESTreeNodeMap.get(res.soleUse),

it seems this is intentional.

Is this good? What do you think?

[JoshuaKGoldberg]

FWIW I'd lean towards reporting on the type parameter itself. This behavior confused & surprised me at first.