How to remove all child nodes of an XmlElement, but keep all attributes?

Question

How to remove all child nodes of an XmlElement, but keep all attributes?

Note, that XmlElement.RemoveAll also removes all attributes. What is a clean, elegant and well-performing way to remove all child nodes? In other words, what is best-practice here?

Gasholder

Answer 1

For a truly efficient solution:

e.IsEmpty = true;

is your fastest and simplest option. It does exactly what you requested: all inner text and nested elements are discarded, while attributes are retained.

Phosphorite

This is the solution I have been waiting for! I verified using Reflector that this property indeed does exactly what is requested. – Gasholder

Answer 2

Would this solution not be simpler?

while(e.FirstChild != null)
    e.RemoveChild(e.FirstChild);

Oratorical

That's kind of what I did. I copied what XmlNode.RemoveAll does. – Gasholder

removeAllChildren in the framework would be simpler. – Limb

Answer 3

Option 1 Use elem.InnerXml = ""; The full working code if you need this:

    var doc = new XmlDocument();
    doc.LoadXml("<x a1='a' a2='b'><child1/><child2/></x>");
    var elem = doc.DocumentElement;

    Console.WriteLine(elem.OuterXml);
    Console.WriteLine("HasAttributes " + elem.HasAttributes);
    Console.WriteLine("HasChildNodes " + elem.HasChildNodes);

    elem.InnerXml = "";
    Console.WriteLine(elem.OuterXml);
    Console.WriteLine("HasAttributes " + elem.HasAttributes);
    Console.WriteLine("HasChildNodes " + elem.HasChildNodes);
    Console.ReadLine();

Detailied information what InnerXml do:

    public override string InnerXml
    {
      get
      {
        return base.InnerXml;
      }
      set
      {
        this.RemoveAllChildren();
        new XmlLoader().LoadInnerXmlElement(this, value);
      }
    }

There could be performance issue on LoadInnerXmlElement but because we have empty string it should not be big, because most time will take this method:

internal XmlNamespaceManager ParsePartialContent(XmlNode parentNode, string innerxmltext, XmlNodeType nt)
    {
      this.doc = parentNode.OwnerDocument;
      XmlParserContext context = this.GetContext(parentNode);
      this.reader = this.CreateInnerXmlReader(innerxmltext, nt, context, this.doc);
      try
      {
        this.preserveWhitespace = true;
        bool isLoading = this.doc.IsLoading;
        this.doc.IsLoading = true;
        if (nt == XmlNodeType.Entity)
        {
          XmlNode newChild;
          while (this.reader.Read() && (newChild = this.LoadNodeDirect()) != null)
            parentNode.AppendChildForLoad(newChild, this.doc);
        }
        else
        {
          XmlNode newChild;
          while (this.reader.Read() && (newChild = this.LoadNode(true)) != null)
            parentNode.AppendChildForLoad(newChild, this.doc);
        }
        this.doc.IsLoading = isLoading;
      }
      finally
      {
        this.reader.Close();
      }
      return context.NamespaceManager;
    }

Option 2 Following code:

    XmlNode todelete = elem.FirstChild;
    while (todelete != null)
    {
        elem.RemoveChild(elem.FirstChild);
        todelete = elem.FirstChild;
    }

About performane. Let's look into XmlElement.RemoveAll() it is:

public override void RemoveAll()
{
  base.RemoveAll();
  this.RemoveAllAttributes();
}

Where base.RemoveAll() is exactly:

[TargetedPatchingOptOut("Performance critical to inline across NGen image boundaries")]
public virtual void RemoveAll()
{
  XmlNode oldChild = this.FirstChild;
  for (; oldChild != null; {
    XmlNode nextSibling;
    oldChild = nextSibling;
  }
  )
  {
    nextSibling = oldChild.NextSibling;
    this.RemoveChild(oldChild);
  }
}

So it is same as I wrote above

Dank

I'd have reservations regarding performance for this one. It also feels hacky. I'd like to deal with the XML document on an AST level, not on a string level. – Gasholder

I added detailed explanation – Dank

Ok thanks! I'll put this into an extension method and the problem is now solved well for all times. – Gasholder