Rational

Keep thinks simple and stupid.

Adding a redis service makes thinks more complicated

This is a try to use plain python utilities to provide channels layer functionality without configuration

idea

we leverage multiprocessing for providing a channel layer. As multiprocessing doesn't play always nice with async we use a per layer a Thread to serialize the internal requests. Per default the default mp_context is used for creating the manager for multiprocessing synchronization It may be set manually to "spawn" in case of an non python asgi server with multiple process workers

Usage

CHANNEL_LAYERS = {
"default": {
        "BACKEND": "channels_multiprocessing.MultiprocessingChannelLayer"
    }
}

with explicit context

CHANNEL_LAYERS = {
"default": {
        "BACKEND": "channels_multiprocessing.MultiprocessingChannelLayer",
        "CONFIG": {
            "mp_context": "spawn",
        },
    }
}

Note: all options of BaseLayer are supported (e.g. capacity)

State

tests passed

TODO

• documentation
• investigate aioprocessing