Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

Table of Contents

Status

Current state: "Under Discussion" Accepted

Discussion thread: https://lists.apache.org/thread/5fmgl51x7h3gsx9mn7xq0n28j700rys0

Voting thread:https://lists.apache.org/thread/jgtzlb4l60hs6t3o67x53w8gonk2ffvo

PR: https://github.com/apache/kafka/pull/12685Please keep the discussion on the mailing list rather than commenting on the wiki (wiki discussions get unwieldy fast).

JIRA: 

Jira
serverASF JIRA
serverId5aa69414-a9e9-3523-82ec-879b028fb15b
keyKAFKA-14945

Motivation

Currently, we use Serializer#serializeSerializer#serialize(String, Headers, T)  in KafkaProducer#doSendKafkaProducer#doSend(ProducerRecord, Callback)  to serialize key &and value. First, we first call Serializer#serialize Serializer#serialize(String, Headers, T) to convert  T  into  byte[] and , then calluse Utils#wrapNullable(byte[])  to convert byte[] into ByteBuffer, and finally append ByteBuffer to DefaultRecordwrite ByteBuffer into MemoryRecordsBuilder through DefaultRecord#writeTo(DataOutputStream, int, long, ByteBuffer, ByteBuffer, Header[]).

. Why don't we add the default method a serializeToByteBuffer(String, Headers, T) method to Serializer, and then use Serializer#serializeToByteBuffer(String, Headers, T) and use it in KafkaProducer#doSend(ProducerRecord, Callback)? If T is an instance of ByteBuffer or T is based on ByteBuffer, it will we would reduce a lot of memory allocation and memory copying.

Additionally, I plan to ultimately replace byte[] with ByteBuffer in Serializer: in 3.x versions, both Serializer#serializeToByteBuffer() and Serializer#serialize() exist. Starting from version 4.0, Serializer#serializeToByteBuffer() will be removed and its implementation will replace Serializer#serialize(), which will return ByteBuffer by default, and Serializer#serialize() will not modify the input ByteBuffer.


Public Interfaces

We propose adding default method Serializer#serializeToByteBuffer(String, T), Serializer#serializeToByteBuffer(String, Headers, T) and Partitioner#partition(String, Object, ByteBuffer, Object, ByteBuffer, Cluster):

Class/InterfaceMethod
Serializer
default ByteBuffer serializeToByteBuffer(String topic, T data) {
return wrapNullable(serialize(topic, data));
}

default ByteBuffer serializeToByteBuffer(String topic, Headers headers, T data) {
return wrapNullable(serialize(topic, headers, data));
}
ByteBufferSerializer
/** 
* Note that this method will modify the position and limit of the input ByteBuffer.
*
* @param topic   topic associated with data
* @param data    typed data
* @return serialized ByteBuffer
*/
@Override
public ByteBuffer serializeToByteBuffer(String topic, ByteBuffer data) {
if (data == null) {
return null;
}

    // Consider that ByteBuffer#wrap(byte[]) return a ByteBuffer that does not need to call flip().
if (data.position() > 0) {
data.flip();
}
return data;
}

@Override
public ByteBuffer serializeToByteBuffer(String topic, Headers headers, ByteBuffer data) {
return serializeToByteBuffer(topic, data);
}
Partitioner
default int partition(String topic, Object key, ByteBuffer keyBytes, Object value, ByteBuffer valueBytes, Cluster cluster) {
return partition(topic, key, toNullableArray(keyBytes), value, toNullableArray(valueBytes), cluster);
}

Proposed Changes

There are the following changes:

...

Compatibility, Deprecation, and Migration Plan

  • Compatibility

    • We

    This proposal has no compatibility issues, we
    • just add default methods Serializer#serializeToByteBuffer(String, T), Serializer#serializeToByteBuffer(String, Headers, T) and Partitioner#partition(String, Object, ByteBuffer, Object, ByteBuffer, Cluster) which is compatible with the existing Serializers.

    • The impact on ByteBuffer offsets is consistent after calling ByteBufferSerializer#serialize(String, ByteBuffer) and ByteBufferSerializer#serializeToByteBuffer(String, Headers, T).

Rejected Alternatives

There is no alternative.