Socket.io server in elixir

NobbZ

As we are currently evaluating server side technologies and I’m in the position to push elixir in my company here, I want to re-implement the current functionality of one of our backend modules in elixir, showcasing its capabilities.

I have some restrictions when doing so though of which I can solve the most, but one remains:

I am not allowed to break the current client software, which uses socket.io for real time communication with the backend among the usual HTTP-RESTful-API.

Is anyone of you aware of a server-side socket.io implementation in elixir (or erlang) that nicely integrates with Phoenix or Raxx?

I doubt beeing able to implement one on my own in the time I have.

10 comments

/phoenix #raxx #websockets

5 7003 10

2024-08-20 23:37:14 UTC

Showing Posts 1 to 10

Show Best Posts
Show All (oldest first)
Show All (newest first)

allyraza

I am not aware of elixir implementation of socketio server, But I do know of an erlang implementation which I don’t think is actively being developed

Post #1

chrismccord

Creator of Phoenix

Phoenix 1.4 included our rewrite of the channels layer, so you should be able to implement the Phoenix.Socket behavior and speak the socket.io contract. Check the Phoenix.Socket source to see how we implement multiplexed channels on top of the behaviour.

Post #2

NobbZ

Last commit was many years ago, so I fear it is to old for what I need. My front-end GUI told me that I need to be able to communicate with a 2.0 client.

Neither him nor me are sure if that is a protocol version or the clients version.

I’m also able to tell, that on the server side currently flask-socketIO is used in version 3.0.x.

If I only knew any details about that contract or they were properly specified in terms of a human readable document rather than untyped JavaScript as a reference implementation…

But as I have the python sources of the currently used server side library, perhaps I am able to re-implement the used subset for a start and extend on that…

A single week is very little time for re-implementing one of the transport protocols… Logic in that module is asymptotic to zero. Mainly its querying a database on one endpoint (and converting the database output to the needed JSON structure) and relaying messages from AMQP to the client and translating those system events to a simplified form that only contains category and label.

Post #3

nonblockio

Hi folks, I meet the same problem recently Is there any solution?

Post #4

namxam

Yeah, I am coming back to this thread from time to time. I have a project, where we are hooking into unreal and using a socket io plugin for communication. Unfortunately, the team has not that much experience in this topic and is happy that it currently works… I would love to migrate the server side to elixir but need a socket.io compliant server api. With covid-19 budget is extremely limited and i haven’t found time to invest into research

Post #5

namxam

Well, our situation got worse and we have various socket.io version requirements in different projects. I would love to give this a try and keep you updated about my progress. But right now I am not really sure where to get started. I guess I will have a look at the Phoenix.Socket behaviour. If anyone else might have a tip or hint I would be quite happy

Post #6

kartheek

Take a look at Phoenix.Socket.Transport — Phoenix v1.8.8 and you should be implementing socket.io protocol on it.

Below article implements binary WebSocket as Socket Transport -

Phoenix Socket expect heartbeat every 60 seconds - I think you have to figure out something on client side.

If phoenix socket doesn’t work out - may be you have to use cowboy web socket.

I have a question - why do you want to use elixir/erlang for socket.io ?

Post #7

namxam

Actually, for backwards compatibility, while being able to move forward and only having to maintain a single code base.

It is a kind of legacy situation. We have multiple projects which depend on different versions of a socket.io plugin. They targets socket io protocol v2, v3, and v4. Each has its own quirks and bugs. And there is no easy path to up- or downgrade those plugins to a single version.

My hope is, to have a minimal implementation of the socket.io protocol (or the various versions) which just suffices to provide those apps with what they need. And they only use a very tiny part of socket.io features.

Post #8

ZaneH

I looked into doing the work for this over a few days but decided it’s too much for me. I was also needing a way to use socket.io with Phoenix to use the Streamlabs Socket API. For now, I have that part of the codebase in a separate Node.js service.

Post #9

proxima

I’ve just wrapped up a spec-compliant engine.io implementation in Elixir, and am going to tackle socket.io here shortly. At my current place of work, integration with Plug.Router is plenty, but I’m curious to hear from others if it’s worth the effort to go a few steps further:

Is supporting all versions of engine.io / socket.io important to you?
Is Phoenix integration important to you?
Would mapping the Phoenix Channels API to socket.io’s “namespaces” be useful? They seem more or less analogous. By this I mean socket.io client-side js interoperability with server-side Phoenix channels.