Sorry for the delay I was on holidays.

I didn't think about this way of using PURE.
Today it works only when the template root is a DOM node which has always a parent.

Would you mind giving me more information about what you are expecting to do after the autoRender is executed?
I addition I guess IE would leek memory (bleed would be a better term) with such in-memory HTML generation.

Cheers,
Mic

There's a typo in my example above. "$p(d)" should be "$(d)"

I expected "d" to now have this innerHTML:

<div class="who">hello</div>

The current engine (as I understand it) expects me to have the template already in the DOM into which I render content. I'd like to not have any placeholder elements and only add the elements (with their rendered content) when I need to.

I got around it with this method (using jQuery):

var generateItem = $(item_template).compile(item_directives, false);
function addItem(data) {
    $(generateItem(data)).appendTo("#items");
}

That usage of compiled functions (that actually return strings) isn't really documented, but it works fine. If that detail is hidden from the users on purpose, and may break in future versions, please let me know :)

The compile returns a function.
And running this function on a JSON will give an HTML string.

In version 1, the compile was creating a function somewhere with a reference.
So almost the same.

I cannot guarantee what will be the future of the library, but this shouldn't change anytime soon.
On the other hand, I'll check your example. It would be interesting to know if we can replace internally(cross-lib) an in-memory node when no parent is available.

Cheers,