Using C# regular expressions to remove HTML tags

Question

How do I use C# regular expression to replace/remove all HTML tags, including the angle brackets? Can someone please help me with the code?

Answer 1

The correct answer is don't do that, use the HTML Agility Pack.

Edited to add:

To shamelessly steal from the comment below by jesse, and to avoid being accused of inadequately answering the question after all this time, here's a simple, reliable snippet using the HTML Agility Pack that works with even most imperfectly formed, capricious bits of HTML:

HtmlDocument doc = new HtmlDocument();
doc.LoadHtml(Properties.Resources.HtmlContents);
var text = doc.DocumentNode.SelectNodes("//body//text()").Select(node => node.InnerText);
StringBuilder output = new StringBuilder();
foreach (string line in text)
{
   output.AppendLine(line);
}
string textOnly = HttpUtility.HtmlDecode(output.ToString());

There are very few defensible cases for using a regular expression for parsing HTML, as HTML can't be parsed correctly without a context-awareness that's very painful to provide even in a nontraditional regex engine. You can get part way there with a RegEx, but you'll need to do manual verifications.

Html Agility Pack can provide you a robust solution that will reduce the need to manually fix up the aberrations that can result from naively treating HTML as a context-free grammar.

A regular expression may get you mostly what you want most of the time, but it will fail on very common cases. If you can find a better/faster parser than HTML Agility Pack, go for it, but please don't subject the world to more broken HTML hackery.

Answer 2

Regex regex = new Regex(@"</?\w+((\s+\w+(\s*=\s*(?:"".*?""|'.*?'|[^'"">\s]+))?)+\s*|\s*)/?>", RegexOptions.Singleline);

Source

Answer 3

I would like to echo Jason's response though sometimes you need to naively parse some Html and pull out the text content.

I needed to do this with some Html which had been created by a rich text editor, always fun and games.

In this case you may need to remove the content of some tags as well as just the tags themselves.

In my case and tags were thrown into this mix. Some one may find my (very slightly) less naive implementation a useful starting point.

   /// <summary>
    /// Removes all html tags from string and leaves only plain text
    /// Removes content of <xml></xml> and <style></style> tags as aim to get text content not markup /meta data.
    /// </summary>
    /// <param name="input"></param>
    /// <returns></returns>
    public static string HtmlStrip(this string input)
    {
        input = Regex.Replace(input, "<style>(.|\n)*?</style>",string.Empty);
        input = Regex.Replace(input, @"<xml>(.|\n)*?</xml>", string.Empty); // remove all <xml></xml> tags and anything inbetween.  
        return Regex.Replace(input, @"<(.|\n)*?>", string.Empty); // remove any tags but not there content "<p>bob<span> johnson</span></p>" becomes "bob johnson"
    }

Answer 4

The question is too broad to be answered definitively. Are you talking about removing all tags from a real-world HTML document, like a web page? If so, you would have to:

• remove the <!DOCTYPE declaration or <?xml prolog if they exist
• remove all SGML comments
• remove the entire HEAD element
• remove all SCRIPT and STYLE elements
• do Grabthar-knows-what with FORM and TABLE elements
• remove the remaining tags
• remove the <![CDATA[ and ]]> sequences from CDATA sections but leave their contents alone

That's just off the top of my head--I'm sure there's more. Once you've done all that, you'll end up with words, sentences and paragraphs run together in some places, and big chunks of useless whitespace in others.

But, assuming you're working with just a fragment and you can get away with simply removing all tags, here's the regex I would use:

@"(?></?\w+)(?>(?:[^>'""]+|'[^']*'|""[^""]*"")*)>"

Matching single- and double-quoted strings in their own alternatives is sufficient to deal with the problem of angle brackets in attribute values. I don't see any need to explicitly match the attribute names and other stuff inside the tag, like the regex in Ryan's answer does; the first alternative handles all of that.

In case you're wondering about those (?>...) constructs, they're atomic groups. They make the regex a little more efficient, but more importantly, they prevent runaway backtracking, which is something you should always watch out for when you mix alternation and nested quantifiers as I've done. I don't really think that would be a problem here, but I know if I don't mention it, someone else will. ;-)

This regex isn't perfect, of course, but it's probably as good as you'll ever need.