simple-observable.js

var simpleObservable = function(value){
    var handlers = new KeyTree([Object, Object, Array]);
    var obs = {
        get: function(){
            Observation.add(this, "value");
            return this.value;
        },
        set: function(value){
            var old = this.value;
            this.value = value;
            queues.enqueueByQueue(handlers.getNode(["value"]), obs, [value, old], function(handler, context, args){
                return {log: ["simpleObservable changed to", args[0], args[1]]};
            }, ["simpleObservable changed to", value, old]);
        },
        value: value
    };
    CID(obs);
    // keyName => queueName => handlers

    canReflect.assignSymbols(obs, {
        "can.onKeyValue": function(key, handler, queueName) {
            handlers.add([key, queueName || "mutate", handler]);
        },
        "can.offKeyValue": function(key, handler, queueName) {
            handlers.delete([key, queueName || "mutate", handler]);
        }
    });

    return obs;
};

the handlers KeyTree houses handlers organized by property then queue. {value: {mutate: [handler1,handler2], notify: [handler3] } }

Observation.add tells any computes that they depend on the value event of this observable.

handlers.getNode(["value"]) return the handlers for the value event grouped by queue

enqueueByQueue will queue those handlers by task. It takes:

• a context in which to call the handlers (obs);
• the arguments to call those functions with ([value,old])
• A function for meta information, including what to log.
• A final argument for "why this was registered"

CID() adds a unique id. I'd like to get rid of this.

can.onKeyValue is what is called by any Observation (compute) binding to this observable. We add the handler in the handlers key tree.