I'm using asyncio.StreamReader to receive data from a remote source. This class has a read() method which returns a bytes object.

Right now I use another bytes object to accumulate data I receive from the reader:

buffer_size = 128
buffer = b""
new_data = await stream_reader.read(buffer_size)

buffer += new_data

Then eventually I extract a payload message from the buffer, process it and get rid of the raw data from which the message was generated:

buffer = buffer[payload_length:]

There are multiple reallocations here, as bytes objects are immutable so every time I append something to the buffer the whole buffer is reallocated.

I understand that there's bytearray, but it's unclear which operations on it are mutating. Would

ba = bytearray()
ba += new_data

avoid any relocations?

Also, can I shrink a bytearray object while retaining its underlying memory buffer, similar to std::vector.clear in C++?

In general, what's the efficient way of manipulating byte buffers which minimises reallocation?