This repo contains an improved version of the avro serializer from
https://github.com/marcosschroh/python-schema-registry-client/. It expects the schema
to be stored in the record itself in order to mimic the behavior of Confluent's Avro SerDe.
It uses Faust's metadata capability inside the Record
class to read the Avro schema
dynamically.
from faust import Record, Schema, Stream
from faust_avro_serializer import FaustAvroSerializer
from schema_registry.client import SchemaRegistryClient
import faust
app = faust.App('myapp', broker='kafka://localhost')
my_topic_name = "my-dummy-topic"
class MyRecordExample(Record):
_schema = {
"type": "record",
"namespace": "com.example",
"name": "MyRecordExample",
"fields": [
{ "name": "foo", "type": "string" },
{ "name": "bar", "type": "string" }
]
}
foo: str
bar: str
client = SchemaRegistryClient("http://my-schema-registry:8081")
serializer = FaustAvroSerializer(client, my_topic_name, False)
schema_with_avro = Schema(key_serializer=str, value_serializer=serializer)
dummy_topic = app.topic(my_topic_name, schema=schema_with_avro)
@app.agents(dummy_topic)
async def my_agent(myrecord: Stream[MyRecordExample]):
async for record in myrecord:
print(record.to_representation())
When the serializer calls the _dumps
method, it searches for the __faust
field inside the
record.
If the serializer finds the field, it is resolving the class and reads the _schema
field
containing the Avro schema.