Question

Consider the following C# code:

using System.Xml.Linq;

namespace TestXmlParse
{
    class Program
    {
        static void Main(string[] args)
        {
            var testxml = 
            @"<base>
                <elem1 number='1'>
                    <elem2>yyy</elem2>
                    <elem3>xxx   <yyy zzz aaa</elem3>
                </elem1>
            </base>";
            XDocument.Parse(testxml);
        }
    }
}

I get a System.Xml.XmlException on the parse, of course, complaining about elem3. The error message is this:

System.Xml.XmlException was unhandled
  Message='aaa' is an unexpected token. The expected token is '='. Line 4, position 59.
  Source=System.Xml
  LineNumber=4
  LinePosition=59

Obviously this is not the real Xml (we get the xml from a third party) and while the best answer would be for the third party to clean up their xml before they send it to us, is there any other way I might fix this xml before I hand it off to the parser? I’ve devised a hacky way to fix this; catch the exception and use that to tell me where I need to look for characters which should be escaped. I was hoping for something a bit more elegant and comprehensive.

Any suggestions are welcome.

If this is a dupe, please point me to the other questions; I’ll close this myself. I am more interested in an answer than any karma gain.

EDIT:

I guess I didn’t make my question as clear as I had hoped. I know the “<” in elem3 is incorrect; I’m trying to find an elegant way to detect (and correct) any badly formed xml of that sort before I attempt the parse. As I say, I get this xml from a third-party and I can’t control what they give me.

Answer 1

I would recommend that you do not manipulate the data you receive. If it is invalid it’s your client’s problem.

Editing the input so it is valid xml can cause serious problems, e.g. instead of throwing an error you may end up processing wrong data (because you tried your best to make the xml valid, but this may lead to different data).

---

[EDIT]
I still think it’s not a good idea, but sometimes you have to do what you have to do.

Here is a very simple class that parses the input and replaces the invald opening tag. You could do this with a regex (which I am not good at) and this solution is not complete, e.g. depending on your requirements (or lets say the bad xml you get) you will have to adopt it (e.g. scan for complete xml elements instead of only the “<” and “>” brackets, put CDATA around the inner text of a node and so on).

I just wanted to illustrate how you could do it, so please don’t complain if it is slow/has bugs (as I mentioned, I would not do it).

class XmlCleaner
    {

        public void Clean(Stream sourceStream, Stream targetStream)
        {
            const char openingIndicator = '<';
            const char closingIndicator = '>';
            const int bufferSize = 1024;
            long length = sourceStream.Length;
            char[] buffer = new char[bufferSize];
            bool startTagFound = false;
            StringBuilder writeBuffer = new StringBuilder();

            using(var reader = new StreamReader(sourceStream))            
            {
                var writer = new StreamWriter(targetStream);

                try
                {
                    while (reader.Read(buffer, 0, bufferSize) > 0)
                    {
                        foreach (var c in buffer)
                        {
                            if (c == openingIndicator)
                            {
                                if (startTagFound)
                                {
                                    // we have 2 following opening tags without a closing one                                
                                    // just replace the first one
                                    writeBuffer = writeBuffer.Replace("<", "&lt;");

                                    // append the new one
                                    writeBuffer.Append(c);
                                }
                                else
                                {
                                    startTagFound = true;
                                    writeBuffer.Append(c);
                                }
                            }
                            else if (c == closingIndicator)
                            {
                                startTagFound = false;
                                // write writebuffer...
                                writeBuffer.Append(c);
                                writer.Write(writeBuffer.ToString());
                                writeBuffer.Clear();
                            }
                            else
                            {
                                writeBuffer.Append(c);
                            }
                        }
                    }
                }
                finally
                {
                    // unfortunately the streamwriter's dispose method closes the underlying stream, so e just flush it
                    writer.Flush();
                }                
            }
        }

To test it:

var testxml =
            @"<base>
                <elem1 number='1'>
                    <elem2>yyy</elem2>
                    <elem3>xxx   <yyy zzz aaa</elem3>
                </elem1>
            </base>";

            string result;

            using (var source = new MemoryStream(Encoding.ASCII.GetBytes(testxml)))
            using(var target = new MemoryStream()) {

                XmlCleaner cleaner = new XmlCleaner();
                cleaner.Clean(source, target);

                target.Position = 0;
                using (var reader = new StreamReader(target))
                {
                    result = reader.ReadToEnd();
                }
            }

            XDocument.Parse(result);

            var expectedResult = 
                @"<base>
                <elem1 number='1'>
                    <elem2>yyy</elem2>
                    <elem3>xxx   &lt;yyy zzz aaa</elem3>
                </elem1>
            </base>";
            Debug.Assert(result == expectedResult);