Could Message provide payload_mut?

[frankmcsherry]

I'm not a Kafka expert, but looking at the various bits and pieces, it seems possible that the Message trait could provide a

    fn payload_mut(&mut self) -> Option<&mut [u8]>;

Looking at the RdKafka::Message docs, it looks like the payload() method returns a void* without a const modifier (unlike the key_ptr() method which is const void*). My C++ is not super-strong, but my sense is that this signature means that one can safely mutate the results of payload().

The motivation might be limited, but it saves a copy for me (Abomonation does deserialization using in-place pointer correction). If it is a no-brainer, then great! If it is weird and mysterious and scary, I can understand that too.

[thijsc]

Actually this library wraps the C interface, so this is the relevant doc: https://docs.confluent.io/2.0.1/clients/librdkafka/structrd__kafka__message__t.html

Otherwise I have no idea if this is safe, the docs don't make that clear. We should probably ask over at librdkafka.

[benesch]

The official word is in:

There are currently no constraints to modifying the per-message key/payload as long as honouring the key/payload lengths, but this should be considered and implementation detail and there are no formal guarantees that this will continue to be the case in future versions.

In practice I don't see a reason why this would ever change though.