Adt discriminator in non-first position

[codesurf42]

Snippet from output of sbt clean jsoniter-scala-macrosJVM/test:

[info] - should serialize and deserialize case class ADTs using discriminator (1 millisecond)
[info] - should serialize and deserialize case class ADTs using discriminator in different positions *** FAILED *** (4 milliseconds)
[info]   "...ype":"AAA","a":1},{"[type":"BBB","a":1},{"type":"CCC","a":1],"b":"VVV"},{"type":..." was not equal to "...ype":"AAA","a":1},{"[a":1,"type":"BBB"},{"a":1,"type":"CCC"],"b":"VVV"},{"type":..." (VerifyingSpec.scala:68)
[info]   org.scalatest.exceptions.TestFailedException:
[info]   at org.scalatest.matchers.MatchersHelper$.indicateFailure(MatchersHelper.scala:392)
[info]   at org.scalatest.matchers.should.Matchers$AnyShouldWrapper.shouldBe(Matchers.scala:7539)

[plokhotnyuk]

@codesurf42 Thanks for your PR!

Please resubmit it with only your changes.

[codesurf42]

refreshed @plokhotnyuk

[plokhotnyuk]

The correct version of your test would be:

    "serialize and deserialize case class ADTs using discriminator in different positions" in {
      verifyDeser(make[List[AdtBase]](CodecMakerConfig.withRequireDiscriminatorFirst(false)),
        List(AAA(1), BBB(BigInt(1)), CCC(1, "VVV"), DDD),
        """[{"type":"AAA","a":1},{"a":1,"type":"BBB"},{"a":1,"type":"CCC","b":"VVV"},{"type":"DDD"}]""")
      verifyDeser(make[List[AdtBase]](CodecMakerConfig
        .withRequireDiscriminatorFirst(false)
        .withDiscriminatorFieldName(_root_.scala.Some("t"))),
        List(CCC(2, "WWW"), CCC(1, "VVV")), """[{"a":2,"b":"WWW","t":"CCC"},{"a":1,"t":"CCC","b":"VVV"}]""")
    }

The CodecMakerConfig.withRequireDiscriminatorFirst(false) option changes only decoding part of the generated codec, and the encoding part always write the discriminator field in the first position.

[plokhotnyuk]

It could be great degradation of performance for deeply nested ADTs in parsers that do not use intermediate AST representation (like zio-json, borer, etc.) when the discriminator field appears in the last position. So for security reasons the discriminator field allowed only in the first position by default. If you have only trusted input in your system or only non-nested ADTs then you can use CodecMakerConfig.withRequireDiscriminatorFirst(false) without worrying about possible DoS/DoW attacks.

[codesurf42]

Thanks for quick reactions.

It could be great degradation of performance

It is an understandable reason. How about adding something in readme - about the chosen default for the first position and how someone can change it? I can add it in this PR, somewhere around - An optional name of the discriminator field for ADTs - wdyt?