[linter] unnecessary_implements

[FMorschel]

Consider the following code:

abstract class A {}

mixin class MixinClass implements A {}

class C with MixinClass implements MixinClass, A {}

abstract class C2 implements MixinClass, A {}

class D extends A implements A {}

With C using MixinClass as a mixin, there is no need for either implements. I'm less sure about the warning on A since it can be breaking if the author removes the mixin and still wants to implement A, but the same class showing up twice on the declaration is completely unnecessary.

For D, we already trigger implements_super_class (docs).

CC @lrhn, @bwilkerson

[bwilkerson]

I'll note that the implements_super_class diagnostic is a compile time error defined by the specification.

There's also a compile time error specified for repeating a type in the implements clause:

class A {}

class B implements A, A {} // The second `A` is flagged by the `implements_repeated` diagnostic

The suggested diagnostic certainly appears to be consistent with some of the compile time errors.

It isn't clear whether this should be a lint or a warning. I don't see any way to have false positives, and it does seem unlikely that anyone wants code that implements the same interface multiple times.

[lrhn]

It really shouldn't be a spec error. I think the analyzer is perfectly capable of giving reasonable warnings for unnecessary code, without the spec getting in its way.

It is unnecessary code to implement something you would have as a superinterface anyway. A warning is in order for that.

(Writing the same interface twice on the same syntactic implements is just so obviously unnecessary, that there was no excuse.)