Wunderkafka

Added

- pre-configured consumers and producers for avro-value/string-key and JSON-value/string-key

Fixed

- avro: disallow deriving schemas with names which are not valid [avro names](https://avro.apache.org/docs/1.11.1/specification/#names)

Added

- Support for `confluent`-like JSON Schema (de)serialization. This functionality requires `jsonschema` package to be installed.
- Now consumer may return `StreamResult`. It is useful to handle deserialization errors in user's code without raising exceptions in the middle of already consumed batch.

Changed

- `HighLevelSerializingProducer` now accepts separate value and key serializers. This allows usage of mixed serialization protocols for the key and value per producer.
- `Serializer`s now may be configured via specific `Store` instance to handle its own type of schemas correctly (`Store` may be shared or not be shared for some reason).

Fixed

- Nested models are now derived more correctly (e.g., deep nested `Annotated` fields)
- Subject naming strategy for `confluent` is now 'topic-{key|value}' instead of 'topic_{key|value}'

Added

- This changelog file with some info about old versions

Changed

- Updated minimal required version of `dataclasses-avroschema` to `0.53.0`. Prior versions doesn't pass library's test suites against `pydantic V2` and require effort to implement

Fixed

- Nested `BaseModel`-inherited structures are not supported for schema derivation
- `sphinx` docs were broken for some time (since `v0.12.0`, update to `pydantic V2`)
- Config generator of `librdkafka` options now generates code which is more compliant with PEP-8
- Annotations for describing avro Schemas are updated as in `fastavro`

Fixed

- Check if `cmd_init` is set for the HTTP client

Added

- HTTP client may be initialized with `cmd_kinit` argument. When provided, the client starts its own thread to handle kerberos ticket (awaits `kinit` to be present in runtime environment)

Fixed

- Set default timeout for subprocess calls of `klist` and `kinit`.