[jira] [Created] (AVRO-1905) Backward and forward compatible

["Amod Kumar Pandey (JIRA)" <ji...@apache.org>]

Amod Kumar Pandey created AVRO-1905:
---------------------------------------

             Summary: Backward and forward compatible
                 Key: AVRO-1905
                 URL: https://issues.apache.org/jira/browse/AVRO-1905
             Project: Avro
          Issue Type: Bug
          Components: java
            Reporter: Amod Kumar Pandey


I am under the understanding that Avro is both backward and forward compatible (for certain schema changes). But as per my test it is neither backward nor forward compatible.

Maven project with avro

{"namespace": "example.avro",
 "type": "record",
 "name": "user",
 "fields": [
     {"name": "name", "type": "string"},
     {"name": "favorite_number",  "type": "int"}
 ]
}

Producer

public class Producer {
  public static void main(String[] args) throws IOException {
    try (ByteArrayOutputStream outputStream = new ByteArrayOutputStream()) {
      user u1 = user.newBuilder().setFavoriteNumber(1).setName("Amod").build();
      writeBinaryEncodedAvro(u1, outputStream);
      user u2 = user.newBuilder().setFavoriteNumber(2).setName("Pandey").build();
      writeBinaryEncodedAvro(u2, outputStream);
      System.out.println(Arrays.toString(outputStream.toByteArray()));
    }
  }

  static void writeBinaryEncodedAvro(SpecificRecord specificRecord, OutputStream os) throws IOException {
    BinaryEncoder binaryEncoder = EncoderFactory.get().binaryEncoder(os, null);
    @SuppressWarnings("unchecked")
    DatumWriter<SpecificRecord> datumWriter =
        new SpecificDatumWriter<SpecificRecord>((Class<SpecificRecord>) specificRecord.getClass());
    datumWriter.write(specificRecord, binaryEncoder);
    binaryEncoder.flush();
  }

}

Consumer 

public class Consumer {
  public static void main(String[] args) throws IOException {
    byte[] data = {8, 65, 109, 111, 100, 2, 10, 103, 114, 101, 101, 110};
    try (ByteArrayInputStream inputStream = new ByteArrayInputStream(data)) {
      System.out.println(fromBinaryMulti(inputStream));
    }
  }

  static user fromBinary(InputStream is) throws IOException {
    BinaryDecoder binaryDecoder = DecoderFactory.get().binaryDecoder(is, null);
    DatumReader<user> datumReader = new SpecificDatumReader<user>(user.class);
    return datumReader.read(null, binaryDecoder);
  }

  static List<user> fromBinaryMulti(InputStream is) throws IOException {
    List<user> users = new ArrayList<user>();
    BinaryDecoder binaryDecoder = DecoderFactory.get().binaryDecoder(is, null);
    while (!binaryDecoder.isEnd()) {
      DatumReader<user> datumReader = new SpecificDatumReader<user>(user.class);
      users.add(datumReader.read(null, binaryDecoder));
    }
    return users;
  }

}

I changed the schema to 

{"namespace": "example.avro",
 "type": "record",
 "name": "user",
 "fields": [
     {"name": "name", "type": "string"},
     {"name": "favorite_number",  "type": "int"},
     {"name": "favorite_color", "type": "string", "default": "green"}
 ]
}

The following does not work.

Consume code using new schema generated code cannot consume byte array generated by old schema.

Consume code using old schema generated code cannot consume byte array generated with new schema.

Is there a problem in the way I am trying to understand the forward or backward compatibility.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)