runtime_parametrization_circe.md

I am curious about what's the approach you'd follow to allow run-time input/state change the way circe Encoders and Decoders serialize and de-serialize a type.

For example, imagine you want to redact values depending on certain run-time value:

    implicit def sensitiveEncoder[T, RedactedT](
      implicit redactedEncoderEvidence: Encoder[RedactedT],
      encoderEvidence: Encoder[T],
      context: SerdesContext
    ): Encoder[Sensitive[T, RedactedT]] =
      if (context.redactSecrets) {
        redactedEncoderEvidence.contramap[Sensitive[T, RedactedT]] {
          _.redacted
        }
      } else {
        encoderEvidence.contramap[Sensitive[T, RedactedT]] {
          _.value
        }
      }

I was able to do that with Spray Json and Circe (semi automatic derivation) using the same idea: Provide the encoders as functions from the type I want to depend on to Encoder[T]. e.g:

    implicit val barEncoder: SerdesContext => Encoder[Bar] = {
      implicit serdesContext => deriveEncoder[Bar]
    }

And allow seamless derivation with:

implicit def applyContextParam[T, F[_]](
  implicit f: SerdesContext => F[T],
  serdesContext: SerdesContext
): F[T] = f(serdesContext)

( https://github.com/pfcoperez/trycirce/blob/master/src/main/scala/org/pfcoperez/Models.scala )

Note that ^ is just a "demo". The idea here is to allow seamless composition with encoders from different sources.

With this, it is possible to generate the instance of the encoder from the run-time state.

Is this something you've ever tried to do? If so, did you follow a radically different approach?

pfcoperez/runtime_parametrization_circe.md