Inline comments within parameters moved before previous comma

[jakebailey]

Describe the bug

dprint-plugin-typescript version: 0.91.3

Input Code

function foo(arg: number, /** @deferred */ cb: () => void): void {}

interface Foo {
  method<T>(
    arg: { new(...args: any[]): T }, // This is a comment
    arg2: (instance: T) => HTMLElement | undefined,
  ): Disposable;

  method2<T>(
    arg: { new(...args: any[]): T },
    arg2: (instance: T) => HTMLElement | undefined, // This is a comment
  ): Disposable;
}

Expected Output

function foo(arg: number, /** @deferred */ cb: () => void): void {}

interface Foo {
  method<T>(
    arg: { new(...args: any[]): T }, // This is a comment
    arg2: (instance: T) => HTMLElement | undefined,
  ): Disposable;

  method2<T>(
    arg: { new(...args: any[]): T },
    arg2: (instance: T) => HTMLElement | undefined, // This is a comment
  ): Disposable;
}

Actual Output

function foo(arg: number /** @deferred */, cb: () => void): void {}

interface Foo {
  method<T>(
    arg: { new(...args: any[]): T } // This is a comment
    ,
    arg2: (instance: T) => HTMLElement | undefined,
  ): Disposable;

  method2<T>(
    arg: { new(...args: any[]): T },
    arg2: (instance: T) => HTMLElement | undefined // This is a comment
    ,
  ): Disposable;
}

Noticed this when updating DT; 30 some files had instances of this.

This is particularly scary because microsoft/TypeScript#58729 introduces a parameter JSDoc comment that shouldn't be moved around.

[jakebailey]

A bisect pointed to #630; maybe something in deno_ast?

[dsherret]

Not sure how we haven't got a report about this in Deno. It might be something that's only triggered with certain config. I'll take a look tomorrow.

[dsherret]

Caused because swc started incorrectly setting the text range for BindingIdent to not include the type annotation. I'll do a quick fix and then try to fix this in swc.

[dsherret]

Opened swc-project/swc#9290

Temporarily patched this in our wrapper: dprint/dprint-swc-ext#62 (we had to introduce a wrapper around swc's spans because the first index was shifted from 0 to 1 in order to make 0 a magic number)

[jakebailey]

Thank you once again!

[dsherret]

I opened #652 to catch this sooner in the future.