AbstractJackson2MessageConverter does not work with other ObjectMappers from jackson-dataformats-binary

[ajafff]

In what version(s) of Spring AMQP are you seeing this issue?

For example:

3.2.4

Describe the bug

Example uses CBORMapper from jackson-dataformat-cbor

To Reproduce

• Use RabbitTemplate#convertAndSend() to serialize an object and send it to a queue
• Use @RabbitListener to read message from queue and deserialize to object

Expected behavior

Serializing and sending works.
Deserializing does not work because convertBytesToObject converts the payload to UTF-8 string before handing it to the ObjectMapper, causing Can not create parser for non-byte-based source.
I could get deserialization to work if I add a ;charset=ascii to the MimeType, as that causes the raw payload to be handed to the ObjectMapper. Unfortunately this hack causes serialization to fail, because it now calls ObjectMapper#writeValueAsString which is not supported for CBOR.

This is basically asking for it to work out of the box, or being able to override convertBytesToObject (currently private) and a new method convertObjectToBytes (so that I don't have to override and copy the contents of createMessage)

Sample

public class CborMessageConverter extends AbstractJackson2MessageConverter {
    public CborMessageConverter() {
        super(new CBORMapper(), MimeTypeUtils.parseMimeType("application/cbor"));
    }
}

[artembilan]

The logic there on serialization like this:

			if (this.charsetIsUtf8 && this.supportedCTCharset == null) {
				bytes = this.objectMapper.writeValueAsBytes(objectToConvert);
			}
			else {
				String jsonString = this.objectMapper
						.writeValueAsString(objectToConvert);
				String encoding = this.supportedCTCharset != null ? this.supportedCTCharset : getDefaultCharset();
				bytes = jsonString.getBytes(encoding);
			}

So, if it is UTF-8 (technically no MimeType provided for the Jackson2JsonMessageConverter), then we write content as byte[].

On the deserialization we have a different logic:

	private Object convertBytesToObject(byte[] body, String encoding, JavaType targetJavaType) throws IOException {
		if (this.supportedCTCharset != null) { // Jackson will determine encoding
			return this.objectMapper.readValue(body, targetJavaType);
		}
		String contentAsString = new String(body, encoding);
		return this.objectMapper.readValue(contentAsString, targetJavaType);
	}

I guess we can improve that with the this.charsetIsUtf8 flag as well.
But at the same time I don't deny making convertBytesToObject() as protected and have that CborMessageConverter.

Let us know if you are willing to contribute the fix: https://github.com/spring-projects/spring-amqp/blob/main/CONTRIBUTING.adoc !

[ajafff]

thank you for your response @artembilan

I can try to contribute a fix. However, it's not clear to me, what the preferred solution would be.

I'm wondering if we want to use this.objectMapper.getFactory().canHandleBinaryNatively() to determine if we always work with byte arrays instead of doing string conversion/encoding stuff.

[artembilan]

I don't think so.
See its JavaDocs:

    /**
     * Introspection method that higher-level functionality may call
     * to see whether underlying data format can read and write binary
     * data natively; that is, embeded it as-is without using encodings
     * such as Base64.
     *<p>
     * Default implementation returns <code>false</code> as JSON does not
     * support native access: all binary content must use Base64 encoding.
     * Most binary formats (like Smile and Avro) support native binary content.
     *
     * @return Whether format supported by this factory
     *    supports native binary content
     *
     * @since 2.3
     */
    @Override
    public boolean canHandleBinaryNatively() {

So, this one is for case where a property of our data object is a byte[] itself.
Doesn't look like this has anything to do with (de)serialization itself.