[linter] unused_parameter

[stephane-archer]

class Foo {
  static void _a(String a, String b) { // A linter should point out that b is useless here.
    print(a);
  }

  void b(String a) {
    _a(a, "i'm useless no linter told me...");
  }
}

[FMorschel]

Currently, there is a hard no for triggering any non-optional parameters. I'm not sure how the team feels about this.

I'd say that for private executable declarations, it's fine to trigger this too.

CC @bwilkerson

[bwilkerson]

For private functions / methods I think it would be reasonable to produce a diagnostic.

I suspect that the "hard no" is there for purely pragmatic reasons. If there's a public method then there can be overrides (or other methods in the 'override tree' (we don't really have good terminology for that)) that do make use of the parameter, but determining that the parameter is being used in a different implementation of the member is an expensive operation (or impossible if the package is used in code not loaded in the workspace).

Given that non-method functions can't be overridden, it might even be reasonable to produce diagnostics for public functions.

With augmentations we face a similar problem for private methods and functions (the parameter might be used by one of the augmentations). If we don't correctly handle augmentations at this point then we're adding a future cost to do so later, but given that the feature is still open to being changed I'm not sure we'd want to try to support augmentations at this point.

I'll also not that adding this support would be a breaking change. I have no idea how breaking, but it's something to be aware of.

[stephane-archer]

@bwilkerson I would like to mention that this could be an optional lint that can be turned off on specific lines.
This a help programmers to identify that parameters are no longer used because of a previous change, so they can simplify the code.

I don't see any breaking change in that configuration. And having false positives because of overwriting should not be a big deal (the compiler would tell you immediately after the quick fix)

[bwilkerson]

Sorry, I'd gotten the impression from the discussion that this is an existing diagnostic that just wasn't being produced in one case. Is that not true; is this a request for a new lint? I ask because I can't find where it's defined.

[FMorschel]

We have unused_element_parameter (docs) that could trigger here. That is the one I was looking at for my phrase above.

If you write this, v will trigger it:

void _foo({int? v}) {}

[stephane-archer]

@bwilkerson Dart might have a similar linter but none cover the issue here (that I'm aware of). unused_element_parameter could cover this use case if it's updated. I don't know if it's better to have things separate here or to have them together because the lint name match but the description of the lint currently reflect its current use case that doesn't include this one

[bwilkerson]

It's probably better if this is implemented as an extension of the current behavior, though that depends on how much code would be broken by adding this check. Hopefully none, but it's hard to know without trying it. (Of course, if there's no breakage then it isn't a common problem, so it might not need a solution.)