How do you recover the dijit registry after destroying it recursively? - dojo

Question

I am working on an application and was doing something like this:
dojo.ready(
function(){ require['dojo/parser','dijit/registry','dojo/on'],function(.....){
//find a dijit and wrap it in event handling code.});
I was getting an error indicating that dojo was trying to register a widget with an id that was already in use. To solve the problem I entered this line of code:
//before finding the dijit destroy the existing registry.
However, logically this prevents the next line from working because now no widget exists to which I can connect an event. How can I recover the dijit ids?

Answer 1

The best solution is to find out why your code is trying to register a widget with an id that is already in use and change it to not to do so.
The #mschr's solution should work, but I would advise again using it, as it can break your code in many other places and you are likely to spend hours investigating strange behavior of your application.
Anyway, if you are willing to do it that way and automatically destroy widgets with the same ID, do not override registry.add() method. You could do it, but it does not mean, you should do it (especially in programming). Employ dojo/aspect instead to call a function that will destroy the widget with the same ID before registry.add() is called:
require([
"dojo/aspect",
"dijit/registry"
], function(
aspect,
registry
) {
aspect.before(registry, "add", function(widget) {
if(registry.byId(widget.id)) {
registry.byId(widget.id).destroy();
// this warning can save you hours of debugging:
console.warn("Widget with id==" + widget.id + " was destroyed to register a widget with the same id.");
}
return [widget];
});
});
I was myself curious how to accomplish #mschr solution without that override, so I created an jsFiddle to experiment: http://jsfiddle.net/phusick/feXVT/

Answer 2

What happens once you register a dijit is the following; it is referenced by the dijit.registry._hash:
function (widget) {
if (hash[widget.id]) {
throw new Error("Tried to register widget with id==" + widget.id + " but that id is already registered");
}
hash[widget.id] = widget;
this.length++;
}
Now, every now and then you would have a contentpane in which you would put a widget programatically (programatically, hence dojo.parser handles cpane.unload and derefences / destroys parser-instantiated widgets).
When this happens, you need to hook onto some form of 'unload', like, when your call cpane.set('content' foo) or cpane.set('href', bar). Hook is needed to destroy and unregister the instances you keep of widgets - otherwise you would have a memoryleak in your program.
Normally, once an object has no references anywhere - it will get cleaned out of memory however with complex objects such as a widget might be, 'class-variables' often have reference to something _outside _widget scope which flags the widget unsafe to delete to the garbage collector... Once you get this point, you will know to perform proper lifecycles, yet not before the concept is fully understood..
What you could do is to override the dijit.registry with your own handler and have any widgets that are doublets destroyed automatically like so:
// pull in registry in-sync and with global scoped
// accees (aka dijit.registry instead of dj_reg)
require({
async:false,
publishRequireResult:true
}, [
"dijit.registry"
], function(dj_reg) {
dijit.registry.add = function(widget) {
// lets change this bit
if (this._hash[widget.id]) {
this._hash[widget.id].destroy(); // optinally destroyRecursively
this.remove(widget.id)
}
this._hash[widget.id] = widget;
this.length++;
}
});