dojo.parser.parse doesn't always return

I have this bit of code. It is used to update the form after a select element changes. onChange an "aja开发者_运维百科x" call is made and this bit of code takes care of the response.

The first time everything works as expected. However it the dojo.parser.parse fails to return about 50% of the time.

At first it looked like this:

var targetNode = dojo.byId(node);
targetNode.innerHTML = data;
dojo.parser.parse(targetNode);

Then I read something about the objects existing. So I thought that maybe destroying them would help:

if(dojo.byId(node)) dojo.destroy(node);
dojo.create('div', { id: node }, afternode, 'after');
var targetNode = dojo.byId(node);
targetNode.innerHTML = data;
dojo.parser.parse(targetNode);

That didn't help any. What the h3ll is going on? Sometimes it parses some of the elements. Is this a known problem with the dojo.parser?

If you create the dijits declaratively and use dojo.parser.parse to parse them on-the-fly and you specify the ID of the dijit, once you parse the same HTML fragment twice, you'll get an error says that the dijit's ID has been registered.

<div dojoType="dijit.form.Button" id="myButton" />

The cause is that the dijits have not been destroyed yet and you can not reuse the ID. If you don't specify the ID when declaring it, you won't get this error, but you actually have memory leaks.

The correct way is to destroy the dijits before parsing the HTML fragment again. The return value of dijit.parser.parse is an array list that contains the references of all the dijits it parsed from the HTML fragment. You can keep the list and destroy the dijits first.

if (dijits) {
  for (var i = 0, n = dijits.length; i < n; i++) {
      dijits[i].destroyRecursive();
  }
}
dijits = dojo.parser.parse(targetNode);