Wanted to extend hpack encoders or allow different Encoder and Decoders to be used.

I have a library coming out which I made as a drop in replacement for python hpack called cyhpack and I wanted to try developing a better solution for injecting my own Decoder and Encoder which runs ls-hpack under the hood. As of currently, here is my workaround but I was wondering if a better implementation could be implemented such as a factory attribute or an Abstract class could be made to accommodate my version and the pure python one as well?

from __future__ import annotations

from h2.config import H2Configuration
from h2.connection import H2Connection as _H2Connection

from .hpack import Decoder, Encoder


class H2Connection(_H2Connection):
    """
    Contains an overridable version of H2Connection that drops
    python hpack in exchange for cyhpack for faster performance
    and speed.
    """

    def __init__(self, config: H2Configuration | None = None):
        super().__init__(config)
        self.encoder = Encoder()
        self.decoder = Decoder()

Start by reading h2.connection.H2Connection and H2Configuration, then compare how the current encoder and decoder are initialized with the subclass workaround shown in the issue. The work is done when callers can use alternate Encoder and Decoder implementations without maintaining a custom H2Connection subclass, with coverage for the configuration path.