Travis Brown

@Lasering This is looking great! I do think @felixbr is right about keeping configuration distinct via something like a separate `ConfiguredCodec` type, which could in principle live in a separate...

You've confirmed that `derived` as an imported extension method supports the `derives` syntax in ADT definitions? If so I don't necessarily see a problem with requiring an import for `derives`....

@rleibman I've stopped contributing to Scala projects. In my view the community is too broken to deserve anyone's time, but there are some people who still choose to maintain this...

Downsides: no exhaustivity checking, and if you're using e.g. `Some("myType")` and one of your case classes has a `myType` member, you silently lose round-tripability.

I rebased this and polished it up and would merge it except that it really blows up the circe-core jar size (e.g. more than double for 2.12). One other annoying...

@asakaev My main concern about that approach is that it encodes a representation of ADTs that isn't round-trippable in the general case and doesn't match the one used by default...

@asakaev Oh, sorry—I missed that there's both a `tag` and a `value` (not just a discriminator, like e.g. `{"tag": "a", "i": 42}`), so you're right, it will round-trip. The "arbitrary...

This is a bug in the compiler, which you can reproduce without the macro annotation: ```scala import io.circe.Decoder import io.circe.generic.semiauto.deriveDecoder case class Foo[A](data: A) object Foo { implicit def decodeFoo[A](implicit...

In my experience existential types like this are almost always a mistake, and to be honest thinking about the semantics of stuff like `Encoder[Info[_

I'm sympathetic but seem to remember some motivation for the current behavior. Let's wait a bit for some more feedback—