schema-registry avro primitive different than confluent

[raphaelauv]

Version & Environment

docker.vectorized.io/vectorized/redpanda:v22.1.3

confluentinc/cp-schema-registry:7.1.0

What went wrong?

The schema-registry of redpanda do not manage the same way the primitive avro schema than the confluent schema-registry

What should have happened instead?

localhost:9081 is the port of the redpanda schema-registry

curl -X POST -H "Content-Type: application/vnd.schemaregistry.v1+json" \
--data '{"schema": "{\"type\": \"string\"}"}' \
http://localhost:9081/subjects/test-key/versions

give {"id":1}

if I curl http://localhost:9081/schemas/ids/1 -> it give : {"schema":"{\"type\":\"string\"}"}

---

localhost:8098 is the port of the confluent schema-registry

curl -X POST -H "Content-Type: application/vnd.schemaregistry.v1+json" \
--data '{"schema": "{\"type\": \"string\"}"}' \
http://localhost:8098/subjects/test-key/versions

give {"id":1}

if I curl http://localhost:8098/schemas/ids/1 -> it give : {"schema":"\"string\""}

JIRA Link: CORE-931

[raphaelauv]

same problem with aiven karapace -> https://github.com/aiven/karapace/issues/411

[twmb]

Those two schemas are equivalent, it looks like the confluent schema registry normalized the schema. AFAICT, schema normalization is not the default and needs to be opted into with normalize.schemas=true in the schema serializer, or a normalize=true query parameter when POSTing:

https://docs.confluent.io/platform/current/schema-registry/develop/api.html#post--subjects-(string-%20subject)-versions https://docs.confluent.io/platform/current/schema-registry/serdes-develop/index.html#schema-normalization

In Confluent's Schema Registry, normalization was temporarily made the default but then later reverted because some people want to retain doc fields, etc:

https://github.com/confluentinc/schema-registry/pull/687 <- made default https://github.com/confluentinc/schema-registry/pull/1035 <- reverted by 5.2.x

I'm not sure why you're seeing normalization by default with the registry, as everything indicates that is not the default.

[vuldin]

I community user also reported similar issues today when trying to create a schema with the following command (both with and without the ?normalize=true param):

curl -X POST -H "Content-Type: application/vnd.schemaregistry.v1+json" --data '{"schema": "{\"type\":\"record\",\"name\":\"car\",\"fields\":[{\"name\":\"model\",\"type\":{\"type\":\"string\"}},{\"name\":\"make\",\"type\":\"string\"},{\"name\":\"year\",\"type\":\"float\"}]}"}' "http://localhost:8081/subjects/a1-value/versions"

In Confluent:

{
    "type": "record",
    "name": "car",
    "fields": [
        {
            "name": "model",
            "type": "string"
        },
        {
            "name": "make",
            "type": "string"
        },
        {
            "name": "year",
            "type": "float"
        }
    ]
}

But in Redpanda:

{
    "type": "record",
    "name": "car",
    "fields": [
        {
            "name": "model",
            "type": 
                {
                     "type": "string"
                },
        },
        {
            "name": "make",
            "type": "string"
        },
        {
            "name": "year",
            "type": "float"
        }
    ]
}