.debug .NET and BizTalk Rants

Removing XML Namespaces with LINQ

Generally speaking, removing namespaces in an XML is never a good thing...if you need to remove them, you're doing it wrong!

Unfortunately, I happen to need to fix some code that strips the namespace because it had a glaring flaw:

private static XElement RemoveNamespace(XElement e)
{
  return new XElement(e.Name.LocalName,
    e.HasElements ? e.Elements().Select(n => RemoveNamespace(n)) : (object) e.Value,
    e.HasAttributes ? e.Attributes().Select(a => new XAttribute(a.Name,a.Value)) : null);
}

See it? Don't feel too bad if you don't; I didn't see it, either, until I ran the code through the debugger.

The problem is it can't handle mixed mode content like this:

<p>I have an <a href="http://mylol.com">LOLz.</a> Thanxbai!</p>

It becomes:

<a href="http://mylol.com">LOLz.</a></p>

It's funny that in stackoverflow there's this exact same question and a number of answers...most of them wrong. One used XSLT (using the term "obligatory") which is correct but i think is a bit overkill. One used RegEx and failed. And most of them has this same flaw I presented here.

I chimed in:

public static XElement RemoveAllNamespaces(XElement e)
{
 return new XElement(e.Name.LocalName,
    (from n in e.Nodes()
      select ((n is XElement) ? RemoveAllNamespaces(n as XElement) : n)),
    (e.HasAttributes) ? (from a in e.Attributes() select a) : null);
}

Note that I traverse through Nodes() instead of just Elements()...which will go through each text() nodes...and it will have this nice bonus:

• I don't need to set the XElement.Value explicitly (traversing through text() Nodes does that for me)
• I can copy comment nodes