Automatic creation of ByteBuffers for StructSerializable data structures.

[Turbojax07]

When using a the void-returning .pack() function of a Struct object to save data as a ByteBuffer, one of the parameters is the ByteBuffer object to insert data into. I was wondering why the user is being made to create the ByteBuffer rather than having .pack() create it internally and returning the modified ByteBuffer. I understand having it as an option, as it would allow you to pack multiple data structures into one buffer, but I think that having the function at least return the modified ByteBuffer would be somewhat helpful. Is this at all the kind of thing that would be worth implementing, or not?

[ThadHouse]

Passing in the ByteBuffer allows the allocation of the ByteBuffer to be reused the next time pack is called. This is a fairly large performance gain, and generally something we do a lot in order to save memory resources.

[Turbojax07]

I get having it as an option, but for the initial time that .pack() is run, why not let it create the buffer?

[PeterJohnson]

It's easier to reason about how the interface works this way. Also, in general, users don't call pack() directly, so we want to make it easy for implementers to write. StructBuffer exists to provide management of a ByteBuffer big enough to hold the struct data and serde to/from it.

[Turbojax07]

Ah, cool, I didn't know about the StructBuffer class.