Support enum encoding

[jgoday]

From #105.

It allows to encode/decode enums (as sealed traits and case objects) like

   sealed trait Color
   case object Green extends Color
   case object Yellow extends Color

I think that some things needs some revision or maybe a different/better solution:

1 - When decoding a SealedTrait, lexer tries to consumes '{', if it throws an UnsafeJson exception, we try to match an enum value.
2 - Find a better approach to determine if an object is a case object (should work in scala.js too)

def isCaseObject(x: Any): Boolean = x.getClass.getName.endsWith("$")

[fsvehla]

@fommil I’d appreciate your opinion here.

[fommil]

conceptually this is nice but it is too magical for my personal tastes. I'd rather this was its own macro. Firstly it breaks backwards compatibility, and secondly it means that your entire sealed trait encoding can change just because you add a non case object.

[jgoday]

conceptually this is nice but it is too magical for my personal tastes. I'd rather this was its own macro. Firstly it breaks backwards compatibility, and secondly it means that your entire sealed trait encoding can change just because you add a non case object.

I agree, especially regarding the backwards compatibility and the lexer/exceptions part.
It was just an attempt to implement that use case and experimenting with the existing codebase.
Please, remove the PR if necessary.

[fsvehla]

@jgoday If we implement this with another macro, we could keep compatibility and fail compilation if not every member of the sealed hierarchy is a case object — wdyt?

[jgoday]

@jgoday If we implement this with another macro, we could keep compatibility and fail compilation if not every member of the sealed hierarchy is a case object — wdyt?

@fsvehla Sounds good, what should it be called (instead of DeriveJsonDecoder) ?

[fsvehla]

@jgoday DeriveJsonDecoderEnum?

[jgoday]

@fsvehla, I have implemented it like this jgoday@2f7160a

Two different macros (DeriveJsonDecoderEnum/DeriveJsonEncoderEnum) only for the sealed traits/case objects like enums.

Is this a valid approach ? If so, we should implement toJsonAST/fromJsonAST tests and keep running existing annotations with the new encoder/decoder for enums.