Is instantiating XmlDocument expensive?

Question

I have the following function:

public static XmlNode GetXMLNodeFromString(string strXML)
{
    XmlDocument doc = new XmlDocument();
    doc.LoadXml(strXML);
    return doc.DocumentElement;
}

which can be called a million times in a line of code that gets returned from a database call:

while (reader.Read())
{
    myXMLList.Add(GetXMLNodeFromString((string)reader["GMLString"]));
}

Is there a better way than to keep instantiating the xmlDocument for every row? Or is it fine to do so?

I don't want to this:

XmlDocument doc = new XmlDocument();
while (reader.Read())
{
    myXMLList.Add(doc, GetXMLNodeFromString((string)reader["GMLString"]));
}

because in all reality there is a tree of functions that i would have to add it to. Not just GetXMLNodeFromString.

Can i do something like this:

public static class Common
{
    public static XmlDocument doc = new XmlDocument();
    public static XmlNode GetXMLNodeFromString(string strXML)
    {
        doc.LoadXml(strXML);
        return doc.DocumentElement;
    }
}

What should i do?

Answer 1

I recommend XmlReader if you're having performance problems, as it allows you to parse the XML as it comes in, rather than pre-parsing it and creating an in-memory object model that you're going to have to go back through again anyway. It can require restructuring your code a bit, however, as it gives you sequential (as opposed to random) access. If you don't need random access, it's a good choice.

Answer 2

We have found XmlDocument sluggish, but only because we're thrashing it beyond the scope of its design. Instantiating one for every table row however will cause you massive problems. It's a heavyweight class designed for in-place manipulation - which it is very good at - and if you're creating and destroying them faster than a human being could count in their head, you're overworking it.