Package org.acemq.amqp.api

Interface SchemaRegistry

All Known Implementing Classes:
InMemorySchemaRegistry, JdbcSchemaRegistry
public interface SchemaRegistry
Turns a schema into a small integer and back.

Some formats describe themselves and some do not. Avro's bytes and Protobuf's carry no account of what they are, so a reader must already hold the schema the writer used — that is the trade those formats make for being compact, and it is why a registry exists at all rather than being an operational preference.

Format-neutral on purpose. The same schemas have to be reachable from Avro and Protobuf today and from MQTT and STOMP later, so this interface knows about text and identifiers and nothing about serialisation. A registry that spoke one format would have to be rebuilt for the next.

The identifier is small on purpose too. It travels in the header of every single message, so four bytes rather than sixteen is not fastidiousness — and it is the layout Confluent's clients already read, which is what lets messages written here be read by tools nobody here wrote. A registry is free to key its own records however it likes, by UUID or otherwise; what crosses the wire is this integer.

Identifiers must be stable forever. A message published today may be read next year by a consumer looking up the schema it was written with, so an implementation that hands out fresh identifiers on restart makes every message written before the restart unreadable.

  • Method Details

    • idFor

      int idFor(SchemaDefinition schema)
      Parameters:
      schema - the schema a message is about to be written with
      Returns:
      its identifier, registering it if the registry has not seen it. Registering the same schema twice must return the same identifier

    • schemaFor

      SchemaDefinition schemaFor(int id)
      Parameters:
      id - an identifier read from a message
      Returns:
      the schema it stands for
      Throws:
      AceMqException - if the registry does not know it