NodeList.empty() doesn't work in IE on certain nodes

[Eugene Lazutkin]

The NodeList defines a function that safely dispatches child nodes of listed parents. It doesn't work on a number of nodes in IE, because it uses innerHTML = "" trick resulting in exceptions for nodes with read-only innerHTML. See #7890 for details.

Additional problems:

• It is not available for stand-alone nodes unless wrapped with NodeList.
• Its body is very simple, yet involves an overhead of compiling a function on every call.

Proposal:

1. Move this function to dojo/_base/html.js and make it available for stand-alone nodes too.
2. Compile its body only once.
3. Implement it differently for IE.

Possible implementation on IE:

dojo.empty = function(/* Node */ node){
  var c;
  while(c = node.lastChild){
    dojo._destroyElement(c);
  }
  return node; // Node
};

[Eugene Lazutkin]

See #7975 for notes for suggested dojo._destroyElement() improvements.

[Eugene Lazutkin]

Reassigning some core tickets to me as agreed with jburke.

[Eugene Lazutkin]

(In [16227]) Implementing dojo.empty() with unit tests, !strict, refs #7976.

[Eugene Lazutkin]

The next step is the refactoring of NodeList?. It will be tracked in #8282.

[Eugene Lazutkin]

LiuCougar proposes the following algorithm for IE:

1. Clone the node without children: var newNode = node.cloneNode(false);
2. Replace the node: node.parentNode.replaceChild(newNode, node);
3. Kill the node: dojo.destroy(node);

The relevant reference: ​http://blog.stevenlevithan.com/archives/faster-than-innerhtml --- according to it this method is faster than the "innerHTML" method for many things.

I definitely will include this method in the performance testing I plan to do later.

Additionally this method can be useful for dojo.destroy() implementation, and (according to the article) for innerHTML replacement on non-IE browsers.

[James Burke]

Will using cloneNode(false) have an effect on any DOM event listeners on the cloned node?

[Eugene Lazutkin]

Good point, we should test for it, but I suspect it will not honor event handlers. :-(

[Eugene Lazutkin]

(In [16299]) Aliasing dojo.html._emptyNode() to dojo.empty(), !strict, refs #7976.

[Eugene Lazutkin]

Closing this ticket in favor of #8282.

[pavelpenchev]

Using IE7 and dojo 1.5.0 + some dijit components I get the following error "lastChild is null or not an object".

I suggest the following patch to changeset 16227 :

<	for(var c; c = node.lastChild;){ // intentional assignment
<		d.destroy(c);
<	}
---
>	if (node) {
>		for(var c; c = node.lastChild;){ // intentional assignment
>			d.destroy(c);
>		}
>	}

[Eugene Lazutkin]

Replying to pavelpenchev:

Using IE7 and dojo 1.5.0 + some dijit components I get the following error "lastChild is null or not an object".

Dijit components should guard against this in their code --- we do not check against missing nodes intentionally. Closing this ticket for now. Please reopen with more information against affected components.