Hi All! I am looking for some ideas and perspectiv...

# announcements

[Alec Muffett]

Hi All! I am looking for some ideas and perspectives on deserialisation, delegation, and type-safety, please.

For the sake of argument, I am using Protobuf and Square's Wire to serialise data, and to deserialise into Kotlin.

Just like the example at https://github.com/square/wire#protocol-buffers I have a common Dinosaur message which has got a name string 

The first and obvious change in the direction that I want to go, would be to replace the name with a protobuf enum so that I get Dinosaur.STEGOSAURUS (etc) as a nameEnum, which means that the code can find out the soft-type of the message; but that doesn't give me what I want.

The second approach would be to use the protobuf oneof type - and it would essentially do exactly the same soft-type thing - I could use Dinosaur.STEGOSAURUS.foo.bar - but that still does not give me what I want.

What I want is for the deserialisation process of a Dinosaur message to yield a Stegosaurus type; or a TRex or Apatosaurus type, or whatever, so that resulting code can guarantee having an Apatosaurus passed into it. I am presuming that some sort of DinosaurParser.peekType() and DinosaurParser.getApatosaurusOrThrow() will be needed, along with a type-enforcement is/as check.

But that leaves me with the question of delegation so that I can access the other fields of the Dinosaur message, like height, weight, etc.

• what's the best way for me to convert a "Dinosaur-message-with-TRex-soft-type" in the serialisation domain
• into a kotlin-domain strongly-typed-instance-of-TRex that the compiler can check
• but also inherit (delegate?) all the accessors and methods from the underlying Dinosaur message, into the TRex instance?
It would be nice to do:

class TRex(d: Dinosaur) : Dinosaur by d

...however the protobuf compilers are generating Dinosaur classes, not interfaces, so I don't think that option is open to me?

[Alec Muffett]

For the avoidance of doubt: the fields of all Dinosaur messages, etc, will be common across all of TRex, Stegosaurus, Apatosaurus, etc; so it would be redundant and wasteful to have separate Protobuf "messages" for each, plus it would cause proliferation of other handler code that could simply be written in terms of Dinosaur. The goal here is to use Kotlin type-safety for methods which demand one-or-other kinds of Dinosaur.

[Alec Muffett]

It would be nice to not have to use this otherwise cute hack: https://www.jetbrains.com/help/idea/extract-interface.html

[Alec Muffett]

Moving the discussion to Reddit: https://www.reddit.com/r/Kotlin/comments/k5n54z/seeking_perspectives_on_deserialisation/gegqaq6?utm_source=share&utm_medium=web2x&context=3