💅 `noUnusedImports` does not work with using TS `declare const` about biome HOT 5 CLOSED

I can't find any TS docs about it, but it is valid - see some examples from the TS codebase itself https://github.com/search?q=repo%3Amicrosoft%2FTypeScript+declare+const&type=code

declare const is indeed valid and documented. What I am talking about is making cohabit an import and a declare const with the same name.

I took some time to test locally, and if the imported file exists, I get an error:

// @filename a.jts
export {}


// @filename b.js
import * as ns from "./a.js";
//          ^^
//          Import declaration conflicts with local declaration of 'ns'. ts(2440)

declare const ns: {
  f: () => unknown
};

However, if the imported file doesn't exist (your case?), TypeScript doesn't report the error – this is likely a bug from TSC.
Thus, I think it is an error of using an identical name for an import and a declare const.

from biome.

Got it, thanks for helping investigate, I'll also open a bug with TS about the compiler issue.

from biome.

I was not aware of this kind of declaration merging. Do you have any TypeScript resources that document this? This looks wrong to me 🤔

I could suggest using module augmentation instead:

import * as routeModule from "__SENTRY_WRAPPING_TARGET_FILE__";

declare module "__SENTRY_WRAPPING_TARGET_FILE__" {
  export function GET(...args: unknown[]): unknown;
  /* ... */
}

export const GET = [routeModule.GET, 'GET'];

I think it is more idiomatic in TypeScript.

from biome.

I can't find any TS docs about it, but it is valid - see some examples from the TS codebase itself https://github.com/search?q=repo%3Amicrosoft%2FTypeScript+declare+const&type=code

I agree your change feels more idomatic, but I think biome should still account for this edge case.

from biome.

I agree your change feels more idomatic, but I think biome should still account for this edge case.

While I feel the sentiment, I find it very hard to explain something that we can't document. When we create a rule, we always strive to write something that we can properly document, with references too.

If we can't document this edge case, how can we prove that what we are doing is the right way?

from biome.