WIP - Road to v2

[Chuckame]

Bug fixes

• ✅ #166

Improvements

• Revamp the Avro entrypoint completely to reduce the complexity and improve extensibility. Example: Json class in kotlinx serialization. See #163
  • More extensible serialization (allow bytes headers for custom format wrappers like confluent schema registry, or the other avro file standards like single object)
• ✅ #162
• Replace AvroDescriptor by a @CustomSchemaGenerator(MyCustomGenerator::class) for allowing generating custom schemas without using reflection, also usable with custom serializers. See #175 #162
• Caching of schema and other internal processes (record & field namings, aliases, ...). See #159
• Replace AvroName and AvroNamespace by just SerialName, natively working with kotlinx. See #165
• Also improve naming strategy #178
• Reduce the dependencies to the minimum and set some as optional. See #149
• Allow more types for the map key. See #114 #131
• Better contribution guidelines, with semantics checks and more complete code formatter in .editorconfig

New features

• Implement duration logical type
• Handle all the possible schemas as root schema, and not only records. See #80

Non goals

• kotlin multiplatform, as it needs to implement everything ourselves instead of just using the apache avro java library. But still, it'll be taken into account in the new Avro entrypoint design for better extensibility.
• Direct encoding: Same as the previous line, since we would have a better extensibility, selecting different mode of serialization should be easily done, like also writing into json
• Keeping all the exact same features: By example, AvroName and AvroNamespace will be changed to the native single annotation SerialName

Maybe ?

• #141 (maybe, as if we succeed of revamping the Avro entrypoint, we could more easily use the apache avro library standard encoders like single object, json, and more)
• #158 (maybe, because it's not really mandatory, as we currently have the workaround of json)