My current project involves gathering text content from an element and all of its descendents, based on a provided selector.

For example, when supplied the selector #content and run against this HTML:

<div id="content">
  <p>This is some text.</p>
  <script type="text/javascript">
    var test = true;
  </script>
  <p>This is some more text.</p>
</div>

my script would return (after a little whitespace cleanup):

This is some text. var test = true; This is some more text.

However, I need to disregard text nodes that occur within <script> elements.

This is an excerpt of my current code:

// get text content of all matching elements
for (x = 0; x < selectors.length; x++) {
  matches = Sizzle(selectors[x], document);
  for (y = 0; y < matches.length; y++) {
    match = matches[y];
    if (match.innerText) { // IE
      content += match.innerText + ' ';
    } else if (match.textContent) { // other browsers
      content += match.textContent + ' ';
    }
  }
}

It's a bit overly simplistic in that it just returns all text nodes within the element (and its descendants) that matches the provided selector. The solution I'm looking for would return all text nodes except for those that fall within script elements. It doesn't need to be especially high-performance, but I do need it to ultimately be cross-browser compatible.

I'm assuming that I'll need to somehow loop through all children of the element that matches the selector and accumulate all text nodes other than ones within <script> elements; it doesn't look like there's any way to identify JavaScript once it's already rolled into the string accumulated from all of the text nodes.

I can't use jQuery (for performance/bandwidth reasons), although you may have noticed that I do use its Sizzle selector engine, so jQuery's selector logic is available.

Thanks in advance for any help!