Refuse to produce impossible Typeable instances about shapeless HOT 3 CLOSED

Could you clarify what you mean by "it would only produce instances for types derived from nongeneric classes"? Providing instances for (concrete instantiations of) polymorphic types (where well defined) is pretty much the whole point of Typeable, both here and in Haskell.

The key phrase in the above is "where well defined". It's well defined in the case of non-empty container-like things. It's clearly not well defined for function values. The reason your example has slipped through is that Typeable provides a fallback implicit for any type with a ClassTag available, in effect giving Java reflection an apparently parametricity respecting paintjob. That's pretty clearly wrong.

I wonder if dfltTypeable should just go, or is there an honest halfway house here? If it does go are there any additional types for which instances should be provided which aren't currently because they're being covered by the catch all case? At a minimum I suppose we could add an instance for any type with a Generic instance available.

Thoughts?

from shapeless.

Could you clarify what you mean by "it would only produce instances for types derived from nongeneric classes"?

class C
trait T
class GC[A]

There are instances for C and T for free, but GC[x] requires more, perhaps a Generic instance as fallback as you suggest.

from shapeless.

Gotcha ... the default case should only produce Typeable instances for nongeneric classes.

Agreed completely. Annoyingly this can only be done with a macro, but it's the right thing to do ... do you feel like taking a crack at it?

In principle it would be possible to do something for things with a Generic instance if Generic were itself polymorphic (ie. you could have a polymorphic Generic for case class Foo[T](t: T)). That's not supported yet ... it's something I'd like to do eventually.

from shapeless.