In finally get CruiseControl.Net set up to not only automate our build process, but notify us of any breaking changes on checkin, I discovered some issues in using the standard setup and deployment projects. Not wanting to spend a lot of time on it, I decided to look into Wix and see how it could help me streamline things.

It turns out Wix is a whole lot more work. Instead of a simple point and click interface, I’m presented with hand editing the XML file. Boo.

There are a few decent open source projects to give me a GUI interface so I can get my point and click back, but they turned out to either be for Wix 2 or they had some serious bugs in them.

In the end I used Dark to script my MSI file. While this worked out ok, Wix 3 wouldn’t compile the script without me making some changes. I dug through the script it created and I started to wonder how I was going to automate this. The trick is really not to automate it. The trick is to set up your script once and then leave it alone. Modifications are only made when things change in your build. That’s fine, but what about setting it up for the first time? You have to create a GUID for each component (read, every file if you want the files updateable) that gets installed. This can be time consuming if you have quite a few ancilary files that need to be shipped with the release.

Well, I automated that part. I created myself a small console application that will take a folder and a output file name as arguments. The program will traverse this folder and create an include script. It treats every file, directory, and directory of files as individual components.

In case anyone wants to do this but doesn’t want to take the time to write all the code, here it is.

using System;
using System.Collections.Generic;
using System.Linq;
using System.Text;
using System.IO;
using System.Xml;
using System.Xml.XPath;
using System.Reflection;

namespace CreateIncludeScript
    class Program
        static int Main(string[] args)
            if (args.Length < 2 || args.Length > 2)
                return (-1);
            string folder = args[0];
            string outputfile = args[1];
            XmlDocument xmlDoc = new XmlDocument();
            XmlNode root = xmlDoc.AppendChild(xmlDoc.CreateElement("Include", "http://schemas.microsoft.com/wix/2006/wi"));

            XmlNode dir = root.AppendChild(xmlDoc.CreateElement("Directory", "http://schemas.microsoft.com/wix/2006/wi"));
            SetAttr(xmlDoc, dir, "Id", "TARGETDIR");
            SetAttr(xmlDoc, dir, "Name", "SourceDir");

            ProcessDirectory(xmlDoc, dir, folder);

            XmlNode feat = root.AppendChild(xmlDoc.CreateElement("Feature", "http://schemas.microsoft.com/wix/2006/wi"));
            SetAttr(xmlDoc, feat, "Id", "DefaultFeature");
            SetAttr(xmlDoc, feat, "Level", "1");
            SetAttr(xmlDoc, feat, "ConfigurableDirectory", "TARGETDIR");
            ProcessFeatures(xmlDoc, feat, dir);

            return (0);

        static void SetAttr(XmlDocument doc, XmlNode node, string name, string value)
            XmlAttribute attr = node.Attributes.Append(doc.CreateAttribute(name));
            attr.Value = value;

        static void ProcessDirectory(XmlDocument xmlDoc, XmlNode target, string directory)
            string[] files = Directory.GetFiles(directory);
            string[] dirs = Directory.GetDirectories(directory);

            foreach (string file in files)
                FileInfo finfo = new FileInfo(file);

                XmlNode comp = target.AppendChild(xmlDoc.CreateElement("Component", "http://schemas.microsoft.com/wix/2006/wi"));
                string guid = System.Guid.NewGuid().ToString().ToUpper();
                string id = "C_" + guid.Replace("-", "");
                SetAttr(xmlDoc, comp, "Id", id);
                SetAttr(xmlDoc, comp, "Guid", "{" + guid + "}");

                XmlNode fnode = comp.AppendChild(xmlDoc.CreateElement("File", "http://schemas.microsoft.com/wix/2006/wi"));
                SetAttr(xmlDoc, fnode, "Id", "F_" + guid.Replace("-",""));
                SetAttr(xmlDoc, fnode, "Name", finfo.Name);
                SetAttr(xmlDoc, fnode, "KeyPath", "yes");
                SetAttr(xmlDoc, fnode, "DiskId", "1");
                SetAttr(xmlDoc, fnode, "Source", finfo.FullName);

                if (finfo.Extension.ToLower() == ".dll" || finfo.Extension.ToLower() == ".exe")
                    if (IsDotNetAssembly(finfo.FullName))
                        SetAttr(xmlDoc, fnode, "Assembly", ".net");
                        SetAttr(xmlDoc, fnode, "AssemblyManifest", "F_" + guid.Replace("-", ""));
                        SetAttr(xmlDoc, fnode, "AssemblyApplication", "F_" + guid.Replace("-", ""));

            foreach (string dir in dirs)
                DirectoryInfo dinfo = new DirectoryInfo(dir);

                XmlNode dnode = target.AppendChild(xmlDoc.CreateElement("Directory", "http://schemas.microsoft.com/wix/2006/wi"));
                SetAttr(xmlDoc, dnode, "Id", dinfo.Name.Replace("-",""));
                SetAttr(xmlDoc, dnode, "Name", dinfo.Name);

                ProcessDirectory(xmlDoc, dnode, dinfo.FullName);

        static void ProcessFeatures(XmlDocument xmlDoc, XmlNode target, XmlNode root)
            XmlNamespaceManager xmlmgr = new XmlNamespaceManager(xmlDoc.NameTable);
            xmlmgr.AddNamespace("wi", "http://schemas.microsoft.com/wix/2006/wi");
            XmlNodeList comps = root.SelectNodes("descendant::wi:Component", xmlmgr);

            foreach (XmlNode node in comps)
                XmlNode compfeat = target.AppendChild(xmlDoc.CreateElement("ComponentRef", "http://schemas.microsoft.com/wix/2006/wi"));
                SetAttr(xmlDoc, compfeat, "Id", node.Attributes.GetNamedItem("Id").Value);
                XmlNode file = node.SelectSingleNode("wi:File", xmlmgr);
                if (file.Attributes.GetNamedItem("Assembly") != null)
                    SetAttr(xmlDoc, compfeat, "Primary", "yes");

        static bool IsDotNetAssembly(string fileName)
            using (FileStream fs = new FileStream(fileName, FileMode.Open, FileAccess.Read))
                    using (BinaryReader binReader = new BinaryReader(fs))
                            fs.Position = 0x3C; //PE Header start offset
                            uint headerOffset = binReader.ReadUInt32();
                                                         fs.Position = headerOffset + 0x18;
                            UInt16 magicNumber = binReader.ReadUInt16();
                                                         int dictionaryOffset;
                            switch (magicNumber)
                                case 0x010B: dictionaryOffset = 0x60; break;
                                case 0x020B: dictionaryOffset = 0x70; break;
                                    throw new Exception("Invalid Image Format");

                            //position to RVA 15
                            fs.Position = headerOffset + 0x18 + dictionaryOffset + 0x70;

                            //Read the value
                            uint rva15value = binReader.ReadUInt32();
                            return (rva15value != 0);

TableAdapter Connection Strings

* 7/29/2010 Update – I have uploaded a sample project here for those of you still struggling with this. It includes C# and VB.Net projects. Enjoy!

I don’t know how a lot of people handle it, but one thing that always bugged me about table adapters and datasets is the way it handles it’s connection strings. It starts out innocent enough. A new dataset is created with it’s associated table adapter. The connection string is saved to the application settings file and that property is saved in the dataset. Running it on the development machine (or on the same network) is no big deal and just works. But what if you send that application to someone else that has their own SQL Server?

Did anyone at Microsoft actually use this scenario in a production environment? What were they thinking?

In order to run that in a production environment you have to set the connection string in the app.config file. While this may be fine for some people, what about the people like me that do not want my users to access that database? I don’t want them to have that username and password… While the chances of a normal user loading SQL Server Management Studio and logging in are slim, it’s still possible and it’s definitely possible if a user purposely wants to get out of having to do work that day.

What are the options? Well, one option is to use the encryption to encrypt the settings in app.config. For me, this option is not ideal. Reports of those settings getting corrupted are quite high, plus you have to deal with the loading and saving of those settings, which isn’t all that easy to do.

The other option is to take the route I was taking for a while… I had a function that would build me a connection string. Then I could: TableAdapter.Connection.ConnectionString = myLibrary.ConnectionStringFunction()

This was great, until that day came where I was in a hurry and added a few more tables to a form but forgot to set the ConnectionStrings. Whoops.

So I needed a solution that would stop me from having to set those ConnectionString properties, keep my connection string out of the app.config, and be easy to use (i.e. Just Works).

I started out by just giving all my datasets the same connection string. Then on application startup, I tried to change that one application setting. Hmm.. It seems those ConnectionString properties are set to friend and are read only.

Upon further investigation, it seems that there are some events that fire, such as SettingsLoaded. This event fires when the app.config is read and all the settings are loaded. When this event fires, it fires inside the MySettings class. This should allow that property to be changed.

    Private Sub MySettings_SettingsLoaded(ByVal sender As Object, ByVal e As System.Configuration.SettingsLoadedEventArgs) Handles Me.SettingsLoaded
      Me.Item("MyAppConnectionString") = MyLibrary.BuildConnectionString()
    End Sub

This will set the MyAppConnectionString setting to the proper connection string. Now, all table adapters will have an up to date connection string.

So what happens if you want to change the connection string later while the application is still running? Well, there is no way to do that. So it’s time to come up with a way to trick it into updating that property.

In looking at the MySettings class, there is another event called PropertyChanged. We can use this event if we create another setting that can be updated anywhere in the application. First, we create a new string setting that has a User scope (I called mine ConnectionString). This will allow the application to update the setting at any time.

Next, we need to create a function that will update that property with our connection string.

  Public Shared Sub ChangeConnectionString()
    My.Settings.ConnectionString = BuildConnectionString()
  End Sub

Now we can change the events in the MySettings class to look like this.

    Private Sub MySettings_PropertyChanged(ByVal sender As Object, ByVal e As System.ComponentModel.PropertyChangedEventArgs) Handles Me.PropertyChanged
      If e.PropertyName = "ConnectionString" Then
        Me.Item("MyAppConnectionString") = My.Settings.ConnectionString
      End If
    End Sub

    Private Sub MySettings_SettingsLoaded(ByVal sender As Object, ByVal e As System.Configuration.SettingsLoadedEventArgs) Handles Me.SettingsLoaded
    End Sub
  End Class

Now, every time that ChangeConnectionString() is called, the MyAppConnectionString will be updated. The ChangeConnectionString procedure can be changed so that it can accept a string parameter that is the actual connection string. Then you can build a Connection String anywhere and just pass it to that procedure.

XML Parsing in .Net

While working on wxDesktop, which was my first attempt at parsing data from an XML source, I tried to find the best way to read and parse out certain data from an XML document. Before I even really got to the meat of it, I had just had it in my head that I was going to use a DataSet. Why would I do that, you ask? Well, when that’s all you know how to do, that’s usually what you go with. I have never dealt with the XML namespace, mostly because I haven’t had to. Now that I have, I know I will never use the crappy DataSet class for this ever again.

Take the following XML document.

<p>      Sample set name<br />
      5<br />
      4<br />
      6<br />
      1<br />
      20<br />
      54<br />
      10<br />
      19<br />
      17<br />
<p>      Second Sample set name<br />
      10<br />
      12<br />
      14<br />
      2<br />
      4<br />
      6<br />
      8<br />
      16<br />
      18<br />

When you read this document in using the DataSet.ReadXML function like this:
private void readXMLDataSet(string FileName)<br />
{<br />
    DataSet ds = new DataSet();<br />
    ds.ReadXml(&quot;test.xml&quot;);<br />
}<br />

The DataSet will contain 3 tables. It will contain the tables called data, value, and otherdata. Not really what you would think would be in there… It sort of makes sense, except for the fact that all value columns are then lumped into one table and then linked via an ID. While this isn’t too bad, and it would probably be usable, it’s not ideal, nor fun to use.

Now take this piece of code:

private void readXMLNodes(string FileName)<br />
{<br />
    System.Xml.XmlDocument xd = new System.Xml.XmlDocument();<br />
<p>    System.Xml.XmlNode node = xd.SelectSingleNode(&quot;/test/data&quot;);<br />
    System.Xml.XmlNode nameNode = node.SelectSingleNode(&quot;name&quot;);<br />
    System.Xml.XmlNodeList valueList = node.SelectNodes(&quot;value&quot;);<br />
}<br />

This uses the XmlDocument, XmlNode, and XmlNodeList classes and allows you to easily parse out certain parts of the XML document to find exactly what you need. nameNode.InnerText will show “Sample set name” and the valueList will contain every value in the data section. So valueList[0].InnerText will be 5.

Now, you could shove this information into a correctly laid out DataSet, but why bother? Now I see the value in creating a class just to parse certain XML documents. You can set the class up so that it will match the structure as it should be logically laid out and use the class to just access the nodes as needed through functions.

← Previous Page