How to normalize HTML in JavaScript or jQuery?

Tags can have multiple attributes. The order in which attributes appear in the code does not matter. For example:

<a href="#" title="#">
<a title="#" href="#">

How can I "normalize" the HTML in Javascript, so the order of the attributes is always the same? I don't care which order is chosen, as long as it is always the same.

UPDATE: my original goal was to make it easier to diff (in JavaScript) 2 HTML pages with slight differences. Because users could use different software to edit the code, the order of the attributes could change. This make the diff too verbose.

ANSWER: Well, first thanks for all the answers. And YES, it is possible. Here is how I've managed to do it. This is a proof of concept, it can certainly be optimized:

function sort_attributes(a, b) {
  if( a.name == b.name) {
    return 0;
  }

  return (a.name < b.name) ? -1 : 1;
}

$("#original").find('*').each(function() {
  if (this.attributes.length > 1) {
    var attributes = this.attributes;
    var list = [];
    
    for(var i =0; i < attributes.length; i++) {
      list.push(attributes[i]);
    }

    list.sort(sort_attributes);
 
    for(var i = 0; i < list.length; i++) {
      this.removeAttribute(list[i].name, list[i].value);
    }

    for(var i = 0; i < list.length; i++) {
      this.setAttribute(list[i].name, list[i].value);
    }
  }
});

Same thing for the second element of the diff, $('#different'). Now $('#original').html() and $('#different').html() show HTML code with attributes in the same order.

JavaScript doesn't actually see a web page in the form of text-based HTML, but rather as a tree structure known as the DOM, or Document Object Model. The order of HTML element attributes in the DOM is not defined (in fact, as Svend comments, they're not even part of the DOM), so the idea of sorting them at the point where JavaScript runs is irrelevant.

I can only guess what you're trying to achieve. If you're trying to do this to improve JavaScript/page performance, most HTML document renderers already presumably put a lot of effort into optimising attribute access, so there's little to be gained there.

If you're trying to order attributes to make gzip compression of pages more effective as they're sent over the wire, understand that JavaScript runs after that point in time. Instead, you may want to look at things that run server-side instead, though it's probably more trouble than it's worth.

Take the HTML and parse into a DOM structure. Then take the DOM structure, and write it back out to HTML. While writing, sort the attributes using any stable sort. Your HTML will now be normalized with regard to attributes.

This is a general way to normalize things. (parse non-normalized data, then write it back out in normalized form).

I'm not sure why you'd want to Normalize HTML, but there you have it. Data is data. ;-)

This is a proof of concept, it can certainly be optimized:

function sort_attributes(a, b) {
  if( a.name == b.name) {
    return 0;
  }

  return (a.name < b.name) ? -1 : 1;
 }

$("#original").find('*').each(function() {
  if (this.attributes.length > 1) {
    var attributes = this.attributes;
    var list = [];

    for(var i =0; i < attributes.length; i++) {
      list.push(attributes[i]);
    }
 
     list.sort(sort_attributes);

    for(var i = 0; i < list.length; i++) {
      this.removeAttribute(list[i].name, list[i].value);
    }

     for(var i = 0; i < list.length; i++) {
       this.setAttribute(list[i].name, list[i].value);
    }
  }
 });

Same thing for the second element of the diff, $('#different'). Now $('#original').html() and $('#different').html() show HTML code with attributes in the same order.

you can try open HTML tab in firebug, the attributes are always in same order

Actually, I can think of a few good reasons. One would be comparison for identity matching and for use with 'diff' type tools where it is quite annoying that semantically equivalent lines can be marked as "different".

The real question is "Why in Javascript"?

This question "smells" of "I have a problem and I think I have an answer...but I have a problem with my answer, too."

If the OP would explain why they want to do this, their chances of getting a good answer would go up dramatically.

The question "What is the need for this?" Answer: It makes the code more readable and easier to understand.

Why most UI sucks... Many programmers fail to understand the need for simplifying the users job. In this case, the users job is reading and understanding the code. One reason to order the attributes is for the human who has to debug and maintain the code. An ordered list, which the program becomes familiar with, makes his job easier. He can more quickly find attributes, or realize which attributes are missing, and more quickly change attribute values.

This only matters when someone is reading the source, so for me it's semantic attributes first, less semantic ones next...

There are exceptions of course, if you have for example consecutive <li>'s, all with one attribute on each and others only on some, you may want to ensure the shared ones are all at the start, followed by individual ones, eg.

<li a="x">A</li>
<li a="y" b="t">B</li>
<li a="z">C</li>

(Even if the "b" attribute is more semantically useful than "a")

You get the idea.

it is actually possible, I think, if the html contents are passed as xml and rendered through xslt... therefore your original content in XML can be in whatever order you want.