Inefficient error handling on JsonDeserializer

[ccrafael]

Actually in the class org.springframework.kafka.support.serializer.JsonDeserializer when an error happen it is handled like this:

catch (IOException e) {
		throw new SerializationException("Can't deserialize data [" + Arrays.toString(data) +
				"] from topic [" + topic + "]", e);
	}

which I think is very inefficient as each byte of the payload is duplicated as a long array of numbers written as a String in the exception message. In my humble opinion the Arrays.toString(data) should be removed from the message and the original data[] object should be added to the SerializationException class.

[garyrussell]

Good point, but SerializationException is not part of this project, but we could make a sub class.

However, adding the byte[] to the to the exception is moot - because, normally, this is used in conjunction with an ErrorHandlingDerserializer, which adds the raw data to the org.springframework.kafka.support.serializer.DeserializationException anyway.

Not using an ErrorHandlingDerserializer is pointless because Spring can't recover from such situations (and kafka-clients is unaware).

I will convert this to an issue; to remove the Arrays.toString().

[garyrussell]

#2838

[garyrussell]

However, when NOT using an EHD, the user will have no clue as to why the deserialization failed (which was probably the original reason that we included the complete JSON in the exception). So maybe we should retain it in that case.

[garyrussell]

On the other hand, Jackson does report the offset within the data where the problem occurred, so the user could get the data using the command line tools.

[ccrafael]

d'oh!! I didnt pay attention! You are right SerializationException is a kafka class. Anyway ErrorHandlingDerserializer is actually wrapping the exception into a DeserializationException, adding the data, the headers and serializing all using Java so the error with the data in the message already duplicated is also serialized and the data is duplicated in memory again.

[garyrussell]

Regardless of the duplication, I agree the array of numbers was useless; so I have removed it.

https://docs.spring.io/spring-kafka/reference/whats-new.html

(also back ported to 3.0.x, 2.9.x).