How to get numbers in elements' inner text by javascript's regex

Your best bet is to use innerText or textContent to get at the text without the tags and then just use the regex /\d/g to get the numbers.

function digitsInText(rootDomNode) {
  var text = rootDomNode.textContent || rootDomNode.innerText;
  return text.match(/\d/g) || [];
}

For example,

alert(digitsInText(document.body));

If your HTML is not in the DOM, you can try to strip the tags yourself : JavaScript: How to strip HTML tags from string?

Since you need to do a replacement, I would still try to walk the DOM and operate on text nodes individually, but if that is out of the question, try

var HTML_TOKEN = /(?:[^<\d]|<(?!\/?[a-z]|!--))+|<!--[\s\S]*?-->|<\/?[a-z](?:[^">']|"[^"]*"|'[^']*')*>|(\d+)/gi;

function incrementAllNumbersInHtmlTextNodes(html) {
  return html.replace(HTML_TOKEN, function (all, digits) {
    if ("string" === typeof digits) {
      return "" + (+digits + 1);
    }
    return all; 
  });
}

then

incrementAllNumbersInHtmlTextNodes(
    '<b>123</b>Hello, World!<p>I <3 Ponies</p><div id=123>245</div>')

produces

    '<b>124</b>Hello, World!<p>I <4 Ponies</p><div id=123>246</div>'

It will get confused around where special elements like <script> end and won't recognize digits that are entity encoded, but should work otherwise.

You don't necessarily need RegExp to get the text contents of an element excluding its descendant elements' — in fact I'd advise against it as RegExp matching for HTML is notoriously difficult — there are DOM solutions:

function getImmediateText(element){
    var text = '';

    // Text and elements are all DOM nodes. We can grab the lot of immediate descendants and cycle through them.
    for(var i = 0, l = element.childNodes.length, node; i < l, node = element.childNodes[i]; ++i){
    // nodeType 3 is text
        if(node.nodeType === 3){
            text += node.nodeValue;
        }
    }

    return text;
}

var bodyText = getImmediateText(document.getElementsByTagName('body')[0]);

So here there's a function that will return only the immediate text content as a string. Of course, you could then strip that for numbers with the RegExp using something like this:

var numberString = bodyText.match(/\d+/g).join('');