Split large XML files into small files

• Download demo - 33.3 KB
• Download source code from CodePlex

Introduction

Working with large XML files is not always an easy task. I am referring to files of size 25MB and more. An approach for processing such large XML files may be to split the XML document into smaller files for processing. It is a no brainier if you just want to split a file into multiple files, but what if you need this partial file to be accessible by an XML parser or DOM individually? You need to make sure that you have a complete node at the end of your smaller file, and you want to skip to the next node at the beginning of your next file.

Background

This is the continuation of my previous topic on how to deal with large XML documents: Large XML Files Processing and Indexing.

Using the code

Here is an idea of how you can do it. I tested the code with many different XML files and it works for the majority of XML files. You may get an error if your split size is too small. It also depends on your XML formatting.

Try to use the attached XML document as an example. Also, I have attached the results of this process: the files .part1, .part2, .part3.

Here is how we split the file and how it works:

• Run the EXE as in Fig. 1.
• Select a file to split (a large XML file).



Now split the file: get the XML file first. Call ImportXMLDoc(false); … SplitFile();.

/// <summary>
/// Split file based on max size in MB
/// </summary>
private void SplitFile() {
    ImportXMLDoc(false);
    nodePathDic.Clear();
    if (string.IsNullOrEmpty(filePath)) {
        MessageBox.Show("Select XML File to split");
        return;
    }

    FileInfo fi = new FileInfo(filePath);
    double origFileSize = (double)fi.Length;
    numOfNewFiles = Math.Ceiling(origFileSize / maxFileSplitSize);
    string filePart = Application.StartupPath + "/" + fi.Name + 
                      ".part1" + fi.Extension;
    int fileCnt = 1;
    long writeFilePosition = 0;

    using (StreamReader sr = new StreamReader(filePath, Encoding.UTF8)) {
        int pos = 0;
        filePart = Application.StartupPath + "/" + fi.Name + 
                   ".part" + fileCnt + fi.Extension;
        //Read each line in XML document as regular file stream.
        StreamWriter sw = new StreamWriter(filePart, false);

        Regex rx = new Regex(@"<", RegexOptions.Compiled | 
                             RegexOptions.IgnoreCase);
        string nodeName = string.Empty;
        do {

            string line = sr.ReadLine();
            pos += Encoding.UTF8.GetByteCount(line) + 2;
            // 2 extra bites for end of line chars.

            MatchCollection m = rx.Matches(line);
            //Save index of this node into dictionary
            foreach (Match mt in m) {
                nodeName = line.Split(' ').Length == 0 ? 
                           line.Substring(1, line.LastIndexOf('>') - 1) : 
                           line.Split(new char[] { ' ' }, 
                           StringSplitOptions.RemoveEmptyEntries)[0];
                if (!nodeName.Contains("?xml") && 
                    !nodePathDic.ContainsKey(pos + mt.Index)) {
                    nodePathDic.Add(pos + mt.Index, nodeName);
                }
                break;
            }

            sw.WriteLine(line);
            sw.Flush();
            writeFilePosition = sw.BaseStream.Position;

            //If we at the limit of new file let's get 
            //a last node and write it to this file 
            //and create a new split file.
            if (pos > maxFileSplitSize * fileCnt) {
                int lastNodeStartPosition = 0;
                string lastNodeName = string.Empty;
                string ln = string.Empty;
                string completeLastNode = GetLastNode(filePath, 
                       out lastNodeStartPosition, out lastNodeName);

                //Some synchronization. TODO: needs to be optimized 
                //but it works "AS IS"
                do {
                    //Skip rest of the node....
                    ln = sr.ReadLine();
                    if (ln == null)
                        break;

                    pos += Encoding.UTF8.GetByteCount(ln) + 2;
                } while (!ln.Contains(lastNodeName));

                //Get position where we will begin to read again in our 
                //original XML file. We want to skip to the end of last 
                //complete node we wrote to the file.
                long swPosition = (writeFilePosition - 
                                  (nodePathDic.Keys[nodePathDic.Count - 1] - 
                                   lastNodeStartPosition)) + 2;
                sw.BaseStream.Position = swPosition >= 0 ? swPosition : 0;
                sw.Write("\n");
                sw.WriteLine("<!-- End of " + Application.StartupPath + "/" + 
                             fi.Name + ".part" + fileCnt + fi.Extension + ". " + 
                             fileCnt + " out of " + numOfNewFiles + " -->");

                sw.WriteLine(completeLastNode + "\n\n");
                sw.WriteLine(nodePathDic.Values[0].Replace("<", "</"));

                filePart = Application.StartupPath + "/" + fi.Name + 
                           ".part" + (++fileCnt) + fi.Extension;
                sw.Flush();
                sw.Close();

                sw = new StreamWriter(filePart, false);
                sw.WriteLine(nodePathDic.Values[0]);
                sw.WriteLine("<!-- Start of " + Application.StartupPath + "/" + 
                             fi.Name + ".part" + fileCnt + fi.Extension + ". " + 
                             fileCnt + " out of " + numOfNewFiles + " -->");
                sw.Flush();
            }
        } while (!sr.EndOfStream);

        //Clean up...
        sw.Flush();
        sw.Close();
        sr.Close();
        sw.Close();
    }
}

At the end of the run, you should have the files included in a Zip file.

Let’s take a look at the output of this process:

At the end of each file, note the “<!—End of…." comment line and the complete last node. I added this for a visual effect. I can use it later to join the documents together (that would be in my next article).

The next file will start where the last file ended.

Note

Root nodes are at the beginning and at the end of each document. The output XML file should be good to be used in an XML DOM or a tool like XMLSpy.

Enjoy. If you have any questions, post them here or send me an email.

History

• Created on November 20, 2008.
• Jan 09, 2011: I've changed the logic of how to end the file nodes and start new file nodes. This is a more robust version and based on .NET 4.0 and includes xsd.exe to generate the XML file schema.