PydanticSchemaManagerenables you to use Pydantic models as Avro schemas. Like the
RecordNameSchemaManager, this new manager handles schema registration for you. To serialize a message, you simply supply a Pydantic object. The manager will also deserialize messages into Pydantic objects if the message’s schema corresponds to a managed Pydantic model. Overall this feature provides end-to-end type checking of messages from development to production.
New HTTPX support with
kafkit.registry.httpx.RegistryApi, in addition to the existing aiohttp support.
Documentation is now built with the new Rubin Observatory user guide theme and Sphinx configuration.
__named_schemasannotation added by FastAvro is now stripped, in addition to
__fastavro_parsed. Via @hhromic 🎉
Kafkit is now also available on Conda-Forge (feedstock URL: conda-forge/kafkit-feedstock).
py.typed file is now included to advertise type annotations in Kafkit.
Python versions 3.7 and earlier are no longer supported because Kafkit is adopting the
__future__and native support for
importlib.metadata. Kafkit is explicitly tested with Python 3.8, 3.9, and 3.10.
We’ve added a
kafkit.sslmodule to help connect to Kafka brokers over TLS. The associated documentation includes a tutorial for working with the SSL certificates generated in a Kafka cluster managed by Strimzi.
kafkit.registry.manager.RecordNameSchemaManagerprovides a streamlined workflow for serializing Avro messages using Avro schemas that are maintained in your app’s codebase. The manager handles schema registration for you. To serialize a message, you simply need to provide the data and the name of the schema.
kafkit.registry.sansio.CompatibilityTypeEnum helps you write use valid Schema Registry compatibility types.
We’ve significantly improved Kafkit’s packaging and infrastructure:
Migrate packaging metadata from
pyproject.toml(Kafkit continues to be a setuptools-based project).
Tox now runs tasks like tests, in conjunction with the existing Pytest set up.
Pre-commit hooks lint and format the code base.
Code style is now handled by Black (and in the documentation with blacken-docs).
Full support for type annotations!
tox -e typingvalidates Kafkit’s type annotations with Mypy.
We’ve migrated from Travis CI to GitHub Actions.
The documentation now includes a development guide.
RegistryApi.putwas doing a
PATCHbehind the scenes. That’s fixed now.
deletemethods weren’t returning data. That’s fixed now as well.
All of the RegistryApi’s low-level HTTP methods have more thorough unit testing now to avoid these issues in the future.
Initial release of Kafkit!
This release includes the
kafkit.registry package with a working Confluent Schema Registry API client implemented with a sans I/O design.
There are two client implementations.
One is designed for aiohttp users (
kafkit.registry.aiohttp.RegistryClient), and the other is for I/O-free unit testing (
The clients include schema caches so they can be used as both local stores of schemas, in addition to accessors for remote schemas.
The release also includes a suite of Avro message serializers and deserializers that integrate with Confluent Schema Registry and the Confluent Wire Format (