Schema event builder allocate 4kb buffer to write 8 byte `EventHeader`

[678098]

Is there an existing issue for this?

• I have searched the existing issues

Current Behavior

If you log ByteBuffer[] generated by SchemaEventBuilder just before sending, you can see 2 buffers, one of these has limit==8 and capacity==4096. The event header is written to a separate buffer which is not used after this.

16:34:21,151 (pool-4-thread-1) ERROR RequestManager:319 java.nio.HeapByteBuffer[pos=0 lim=8 cap=4096]
16:34:21,151 (pool-4-thread-1) ERROR RequestManager:319 java.nio.HeapByteBuffer[pos=0 lim=228 cap=4096]
16:34:21,153 (pool-4-thread-1) INFO RequestManager:324 Message sent: Request message: {"rId":3,"openQueue":{"handleParameters":{"uri":"bmq://bmq.test.mem.fanout/java-it-5dfa06ed-f58c-40c0-ba70-9de9be753d77","qId":0,"subIdInfo":{"subId":2,"appId":"bar"},"flags":2,"readCount":1,"writeCount":0,"adminCount":0}}}; Timeout duration: PT20S

Expected Behavior

Will be good to use the same buffer for both event header and the event payload.

Steps To Reproduce

Log

BlazingMQ Java SDK Version

0.0.10

Anything else?

No response

[Wassef-Chebbi]

Hi 👋
I’d like to work on this issue.

From my understanding, the current implementation in EventBuilder.build() creates a separate ByteBufferOutputStream for the header, which results in allocating a 4KB buffer to write only the 8-byte EventHeader. This leads to unnecessary memory overhead and an extra buffer in the final ByteBuffer[].

I’m considering replacing this by:

1- Allocating a header buffer with the exact size (EventHeader.HEADER_SIZE, i.e. 8 bytes), instead of using ByteBufferOutputStream
2- Prepending it to the payload buffers returned by bbos.reset()

This would reduce memory usage and avoid the extra 4KB allocation while keeping the existing structure (still using multiple buffers).

Before proceeding, I wanted to confirm:

• Is there any design reason (e.g. scatter/gather I/O optimization with SocketChannel.write(ByteBuffer[])) that requires the header and payload to remain in separate buffers?
• Or would it be acceptable to optimize the header allocation as described above?

Thanks!