crabka-schema-serde

Client-agnostic, Confluent-compatible schema serializers/deserializers for Crabka.

It frames payloads with the Confluent wire format (magic(0x00) | schema_id(4 BE) | body, plus the Protobuf message-index), registers/resolves schemas against a Confluent-compatible Schema Registry, and supports Avro, Protobuf, and JSON Schema via optional feature flags.

Features

crabka-schema-serde = { version = "0.3.2", features = ["avro"] }

How it works

• A serde is constructed for a key/value role (AvroSerde::<T>::value(&cache) / ::key(&cache)) and derives its subject (<topic>-value / <topic>-key via TopicNameStrategy) from the topic passed at (de)serialize time — so one serde instance works across topics (like JVM Kafka's serialize(topic, data)).
• register_subject(topic) interns the local schema into a shared [SchemaCache]; SchemaCache::prewarm() resolves every interned subject's id once at startup — auto-registering (Confluent default) or looking up.
• serialize/deserialize are synchronous: they read ids from the cache. An unknown writer id on the deserialize path triggers a non-blocking background fetch and returns a retriable error until the cache fills.

Example (Avro)

use std::sync::Arc;
use apache_avro::AvroSchema;
use serde::{Deserialize, Serialize};
use crabka_schema_serde::{RegistryClient, SchemaCache, CacheConfig};
use crabka_schema_serde::format::{avro::AvroSerde, SchemaSerializer, SchemaDeserializer, SchemaSubject};

#[derive(Serialize, Deserialize, AvroSchema)]
struct Order { id: String, total: f64 }

# async fn run() -> Result<(), Box<dyn std::error::Error>> {
let cache = SchemaCache::new(RegistryClient::new("http://localhost:8081"), CacheConfig::default());
let serde = AvroSerde::<Order>::value(&cache);
serde.register_subject("orders"); // intern orders-value for pre-warm
cache.prewarm().await?;            // register/resolve the subject id

let bytes = serde.serialize("orders", &Order { id: "o-1".into(), total: 9.5 })?;
let back: Order = serde.deserialize("orders", &bytes)?;
# Ok(()) }

For a Kafka Streams integration (the Serde<T> bridge + runnable per-format pipelines), see crabka-client-streams with its schema-serde feature.

License

Apache-2.0. See the repository LICENSE and NOTICE files.