Extract text and links from HTML using Regular Expressions

Question

I would like to extract text from an html document keeping the links inside it. for example:

From this HTML code

<div class="CssClass21">bla1 bla1 bla1 <a href="http://www.ibrii.com"&gt;go to ibrii</a> bla2 bla2 bla2 <img src="http://www.contoso.com/hello.jpg"&gt; <span class="cssClass34">hello hello</span>

I would like to extract just this

bla1 bla1 bla1 <a href="http://www.ibrii.com"&gt;go to ibrii</a> bla2 bla2 bla2 hello hello

In another post on StackOverflow i have found the RegEx <[^>]*> which allows to extract text by replacing every match with nothing. How can I exclude the anchor tags from the match? It seems that RegEx do not allow inverse matching.

Answer 1

Temporarily encode <a href ...>...</a> into something else, remove all other tags then restore the <a> tags:

// Example in javascript:
string.
    replace(/<a(.*?)>/g,'\0$1\0').
    replace(/<\/a>/,'\1').
    replace(/<[^>]*>/,'').
    replace(/\0(.*?)\0/,'<a$1>').
    replace(/\1/,'</a>');

In the code above I use the NUL and SOH characters (ASCII 0x00 and 0x01) as replacements for <a> tags simply because it is highly unlikely that they would appear in strings. Feel free to replace them with any other character or sequence of characters that would not appear in your string.

From additional comments it appears you're operating in a browser. In which case the browser has already parsed the HTML for you into a nice DOM tree. Use DOM methods to parse through the tree and process it the way you want:

function simpleHTML (domNode) {
    var ret = "";
    if (domNode.nodeType === Node.ELEMENT_NODE) {
        var children = domNode.childNodes;
        for (var i=0;i<children.length;i++) {
            var child = children[i];

            // Filter out unwanted nodes to speed up processing.
            // For example, you can ignore 'SCRIPT' nodes etc.
            if (child.nodeName != 'SCRIPT') {
                if (child.nodeName == 'A') {
                    ret += '<a href="' + child.href + '">' +
                               simpleHTML(child) +
                           '</a>';
                }
                else {
                    ret += simpleHTML(child);
                }
            }
        }
    }
    else if (domNode.nodeType === Node.TEXT_NODE) {
        ret +=  domNode.nodeValue;
    }
    return ret;
}
// serialize the whole document:
var simpleDocument = simpleHTML(document.body);

// serialize a div:
var simpleDiv = simpleHTML(document.getElementById('some_div'));

// filter a html formatted string:
var temp = document.createElement('DIV');
temp.innerHTML = original_string;
simple_string = simpleHTML(temp);

Answer 2

Regular expressions do allow a non-trivial form of negation through lookahead but in this case it would be just good as an excercise because, while I'n no zealot that burns with holy fire every time regexp get mentioned together with HTML, this is really a problem to be solved with a parser.