SignalR/CoffeeScript Hub callback issue

Question

This works fine

class Window.AppViewModel
    message : ko.observable("")
    chatMessages : ko.observableArray()
    canSendMessage : ko.observable(false)
    Window.sHub.addMsg = (data) =>  
        @::chatMessages.push(data)

whereas this

class Window.AppViewModel
    constructor : ->
        @message = ko.observable("")
        @chatMessages = ko.observableArray()
        @canSendMessage = ko.observable(false)
        Window.sHub.addMsg = (data) =>  
            @chatMessages.push(data)

Results in my server side hub recieving server messages but it can't make any callbacks.

What am I missing here??? This is driving me up the wall. Obviously the prototype based solution will work because it's making everything "static" but I'm crippled because I can't write nicely separated code at the moment.

Is there any way to see what the hub is trying to invoke?? The SignalR documentation talks about tracing/logging but never explains how.

Answer 1

All wired client events can be found by adding a small snippet to the "On" anonymous function in jquery.SignalR in the hubProxy prototype

So now it looks something like this

   on: function (eventName, callback) {
            /// <summary>Wires up a callback to be invoked when a invocation request is received from the server hub.</summary>
            /// <param name="eventName" type="String">The name of the hub event to register the callback for.</param>
            /// <param name="callback" type="Function">The callback to be invoked.</param>
            var self = this;

            // Normalize the event name to lowercase
            eventName = eventName.toLowerCase();
            console.log(eventName + " callback was registered on the client");
            $(self).bind(eventNamespace + eventName, function (e, data) {
                callback.apply(self, data);
            });
            self.subscribed = true;
            return self;
        }

And gives you nice output of what client methods can be called back from the server

Also, I learned that CreateHubProxies (reflects through the callbacks you add) is only called once for an entire hub.Start()--So all methods must be added to the hub proxy before you start.

This should definitely be on the github wiki. It makes sense but it isn't obvious.

Edit : It would make sense to allow you to call CreateHubProxies multiple times. People are accustom to writing Knockout.js code in a way where they delay allocating a viewmodel until a certain event happens--if your viewmodel has any SignalR callbacks, this isn't impossible. Having to create all of your viewmodels upfront is a bit of a kludge. You could even keep it pretty lean by keeping a list of proxy members already generated and excluding them.