MVC application deployed with Visual Studio setup project does not behave like an MVC application

This post is for you if your MVC web application works when debugged and works when published using the built-in Visual Studio publish mechanism, but fails when deployed using a Visual Studio web setup project.  You may receive errors in the browser that indicate directory browsing is not enabled/permitted.

The reason is likely because you added 'primary output' from your web application project to the setup project, but failed to include 'content files'.  Content files include the web.config file, which the web application needs for proper operation.  Right click on the setup project node and click Add | Project Output.  Select the web application project from the drop down and select 'Content Files' and then OK.  

Android Development Setup on a Windows 7, 64-bit Environment with TFS

The nice thing about doing Android development is that the tools are all free, allowing you to delve into mobile development without laying out any cash. Getting started with Android development is simply a matter of downloading and installing a handful of tools.  You could start with Google's Android Bundle which includes many of the tools listed below in a single download.  But sometimes the best way to learn a new development platform is by installing each of its components manually.

Below are instructions targeted for a Windows7 64-bit environment and using Team Foundation Server (TFS) for source control.  Note that the software versions mentioned below are current as of January 2013. 

1. Install Java
   
   Install the Java SDK version 7, update 10.  Download the installer from http://www.oracle.com/technetwork/java/javase/downloads/jdk7-downloads-1880260.html.  When installing, choose all the default options.

2. Install Eclipse
   
   There are a handful of Java IDEs available, but these instructions will guide you on installing Eclipse.  It's free and has a large community following.  Get the Juno release installer for mobile development on Windows 7, 64-bit at http://www.eclipse.org/downloads/download.php?file=/technology/epp/downloads/release/juno/SR1/eclipse-mobile-juno-SR1-win32-x86_64.zip.  Extract the files to C:\Program Files.
   
   Once extracted, edit the C:\Program Files\eclipse\eclipse.ini file using WordPad.  Include the lines:
   
   

   -vm
   C:\Program Files\Java\jre7\bin
   

   
   Then save the file.  I suggest WordPad because the eclipse.ini file may come with Unix-style line breaks, something other editors might not handle.  This change allows the Eclipse application to be pinned to the taskbar, as described in http://brianreiter.org/2010/07/15/pin-eclipse-helios-to-windows-7-taskbar.  Because the file is located in a Program Files directory, you might need to launch WordPad as an administrator depending upon your security settings. 
3. Install Android Developer Tools
   
   Eclipse doesn't natively support the development of Android applications.  To do that you'll need to install the developer tools.  To do this, launch Eclipse and select the Help > Install New Software menu item.  Click the "Add" button.  For the name, enter "Android", and for the location, enter https://dl-ssl.google.com/android/eclipse (if for some reason https fails, try http).  At a minimum, check "Android DDMS" and "Android Development Tools".  Choose all the default options thereafter and then "Finish".  If prompted, click OK to allow unsigned content.  Don't restart Eclipse...simply exit.
4. Install the Android SDK
   
   Download the Android SDK from http://developer.android.com/sdk/index.html.  Don't download the entire bundle, because we've already installed Eclipse.  Just download the SDK Tools for Windows installer and launch it.  Choose whether to install for just you, or all users.  Choose all the default options thereafter and then "Finish". 
   
   The Android SDK Manager will be launched automatically by the installer.  This will allow you to download the Android API packages.  The packages you download will depend upon the devices on which you plan to run, the API version for which can be found on your device in Settings > About Phone > Android Version.  Note that other API versions can be installed at any time by re-running the SDK Manager.exe.  Along with the API, install the"Android Support Library" under Extras. 
   
   If you attempt to install all components at once and receive errors, or they continue to be shown as "Not Installed" even after what appeared to be a successful install, try installing one feature at a time.  Occasionally the SDK Manager has trouble installing multiple features at once and will error or silently fail.
5. Setup Android Virtual Device
   
   To test your application in the absence of a physical Android device you'll need a virtual one.  Create a virtual device with characteristics that resemble your physical one by launching the Android Virtual Device Manager.  To do this, launch Eclipse and select Window > Android Virtual Device Manager.  Click "New" to create a virtual device that match your device's specifications.
6. Install the Team Foundation Server plugin
   
   To enable Eclipse to use TFS, launch Eclipse and go to the Help > Install New Software menu item.  Click the "Add" button.  For the name, enter "TFS", and for the location, enter http://dl.microsoft.com/eclipse/tfs.  Check all features.  Choose all the default options thereafter and then "Finish".  Then restart Eclipse.
7. Ignore binaries in source control
   
   Binary files don't belong under source control.  But you have explicitly to tell Eclipse to ignore them, otherwise they'll be added by default.  In Eclipse, go to Window > Preferences > Team > Ignored Resources > Add Pattern..., and add */bin/* (per http://msdn.microsoft.com/en-us/library/gg413275%28v=vs.100%29.aspx).
8. Connect to TFS
   
   Eclipse must be connected to a TFS server in much the same way that Visual Studio is.  Launch Eclipse and go to the Window > Open Perspective > Other... > Team Foundation Server Exploring option.  Click Connect to Team Projects > Servers... > Add, and enter the server name, path, port, and protocol (the last three are typically "tfs", "8080", and "http", respectively).
9. Install device USB driver
   
   In order to connect your device to your desktop, you'll need to install the device specific USB driver.  I happen to have a Hauwei device, the driver for which I found at http://download-c.huawei.com/tcpsdownload/downLoadCenter?category=&flay=software&downloadID=Mzk1MzA=.
10. Enable debugging
    
    Once you get started you're bound to run into issues that you'll want to debug.  To do this, you must enable debugging on both the device and within your Android project.  To enable debugging on the device, go to Settings > Applications > Development, and check "USB debugging".    From within Eclipse project, open the application AndroidManifest.xml, and on the Application tab set Debuggable to "true".

This property is not available on SQL Server 7.0

You're likely getting this error because you're trying to interface with SQL Server 2012 through the version 10.0 (2008) of the SMO library -- SQL is backwards compatible, not forwards compatible.

Ensure that you are compiling against the version (11.0) 2012 SMO libraries (can be found in the installation SDK directory), and that if you include any SMO libraries with your installation that they are also the version 11.0 binaries.

Bulk Inserts with the Entity Framework

A bulk insert is a SQL command that allows you to insert a large number (100s or 1000s) of records for a single table all at once (similar to an Oracle prepared statement).  Without a bulk insert, the records would have to be inserted one at a time and incur the expense of many SQL commands as opposed to just one bulk insert command. 

As of version 5.0 of the Entity Framework (EF), there is still no support for bulk inserts.  If records are to be inserted in bulk, it will have to be done outside the EF.  There are many examples online on how to do this with an EF context and the SqlBulkCopy object.  However, they assume that the entire SQL transaction will consist of nothing but the records in the bulk insert.  In reality, transactions are almost always going to involve inserting records into more than just one table.  Also, you'll likely want to use bulk insert when it improves performance, but continue to use EF for adding/updating other objects because it is so much more convenient.  This article shows how to accomplish this using the EF.

NOTE: consider disabling lazy loading in conjunction with the below technique.  Lazy-loading (combined with the EF "fixup" feature) has serious performance implications, especially when large numbers or objects are involved.

For example, imagine we are creating a business function that inserts 1 object of type Customer and 10000 of type Order.  To get better performance, we'll use a bulk insert for the Order objects:

public void CreateCustomer(int id)
{
    using(var dc = GetContext())    
    {
        var customer = new Customer { Id = id };
        dc.Customers.AddObject(customer);
        IList<order> orders = new List<order>();

        for(int orderId = 1; orderId <= 10000; orderId++)
        {
            // for example simplification, I did not associate order with customer
            var order = new Order { Id = orderId };       
            orders.Add(order);
        }

        var copy = new SqlBulkCopy(dc.Database.Connection.ConnectionString, "Orders");

        // AsDataReader() taken from http://code.msdn.microsoft.com/LinqEntityDataReader
        copy.WriteToServer(order.AsDataReader());
        dc.SaveChanges();
    }
}

This will "work", but it definitely will not be transactional.  Heck, it won't even all occur using the same SQL connection!  This is because SqlBulkCopy is given a connection string, not a connection - so it won't share a connection with the context.  Second, SqlBulkCopy will begin and end its own transaction, as will the call to SaveChanges().  This means that if there is a failure the Customer might be inserted but not the Orders, or vice versa.

To make this transactional we must (unfortunately) apply several reflection and casting hacks in the process of performing the following steps:

• Forcibly open a database connection
• Manually begin and end an entity transaction
• Obtain the current SQL transaction from the entity transaction
• Instantiate a SqlBulkCopy object using the connection and transaction

We would restructure the above method as follows:

public void CreateCustomer(int id)
{
    using(ObjectContext dc = GetContext())
    {
        // force open the connection
        // if we don't explicitly open it first, the EF will open and close it at will
        // making it impossible to reliably obtain one when we need it
        dc.Connection.Open();

        try
        {
            // create a transaction now so that either the entire operation
            // succeeds or nothing succeeds.  Also, we'll need this transaction
            // later when instantiating the SqlBulkCopy object
            using (var tx = entityConnection.BeginTransaction())
            {
                var customer = new Customer { Id = id };
                dc.Customers.AddObject(customer);

                IList<order> orders = new List<order>();

                for(int orderId = 1; orderId <= 10000; orderId++)
                {
                    // for example simplification, I did not associate order with customer
                    var order = new Order { Id = orderId };      
                    orders.Add(order);
                }

                BulkInsert(dc, orders.AsDataReader());
                dc.SaveChanges();
            }
        }
        finally
        {
            Context.Connection.Close();
        }
    }
}

public void BulkInsert(ObjectContext context, IDataReader reader)
{
    var sqlConnection = context.Connection.StoreConnection as SqlConnection;
    if (sqlConnection != null)
    {
        var parserProp = sqlConnection.GetType().GetProperty("Parser", BindingFlags.NonPublic | BindingFlags.Instance);

        if (parserProp != null)
        {
            var parser = parserProp.GetValue(sqlConnection, null);
            var sqltxProp = parser.GetType().GetProperty("CurrentTransaction", BindingFlags.NonPublic | BindingFlags.Instance);
            var currentTransaction = sqltxProp.GetValue(parser, null);
            var sqlTransaction = currentTransaction as SqlTransaction;

            if (sqlTransaction == null)
            {
                // this means that the current transaction is likely an SqlInternalTransaction and
                // that we must obtain the true, underlying transaction from the Parent property
                var parentProp = currentTransaction.GetType().GetProperty("Parent", BindingFlags.NonPublic | BindingFlags.Instance);
                currentTransaction = parentProp.GetValue(currentTransaction, null);
                sqlTransaction = currentTransaction as SqlTransaction;
            }

            if (sqlTransaction != null)
            {
                var sqlBulkCopy = new SqlBulkCopy(sqlConnection, SqlBulkCopyOptions.Default, sqlTransaction);
                sqlBulkCopy.DestinationTableName = typeof(T).Name;
                sqlBulkCopy.WriteToServer(entities.AsDataReader());
            }
        }
    }
} 

This technique is limited in that objects you are bulk inserting are not tracked by the EF.  This means that the EF is unaware of any object that is bulk inserted which means there is the potential for conflicts if the same object is both bulk inserted and inserted via AddObject().  Also, SqlBulkCopy will not use any of the EF's navigational properties because they have no meaning to it.  You must instead ensure all required scalar properties are initialized.

If the limitations are not an issue for you or easy to avoid (which in general they should be), this technique is useful for improving the efficiency of any business function that requires the insertion of large numbers of records.

Never Explicitly Call DateTime.Now

In order to evoke certain scenarios from your application testing, it is often necessary to make your system think the current time is some point in the past or future.  Traditionally, this is done by modifying the Windows system date/time.  But a change like this affects ALL applications, not just your own -- with potentially nasty results.

At the heart of this issue is the fact that accessing the property DateTime.Now is typically scattered everywhere across an application's code base.  Fetching the value in this manner - explicitly - tightly couples your code to the system clock.

A better way would be to treat the system clock as a separate concern.  Instead of accessing DateTime.Now directly, create an interface, called IClock for example, that indirectly retrieves the current time.  By using IClock, objects would be loosely coupled to the system clock.  In fact, a sophisticated implementation if IClock would allow for the current time to be offset by a certain amount, allowing an application to be tested "in the past" or "in the future", all without having to adjust the actual system time.
 

Avoid Custom .NET Exception Boilerplate Code

In order to guarantee a custom .NET exception will work in all scenarios, Microsoft recommends the exception class observe certain conventions.  Specifically, the class should implement certain constructors and serialization methods.  This means each of an application's custom exceptions will likely end up with the same boilerplate code.  If an application has dozens of unique exceptions this can add up to a lot of code duplication, with each exception having to implement the same constructors and serialization code.

This code duplication can be avoided by using a single, generic exception used application-wide.  This solution involves the creation of a single DomainException<T> class with one or more serialization exception details classes:

[Serializable]
public class DomainException<T> : Exception
{
    public DomainException(T exceptionDetails) : base(default(String))
    {
        Details = exceptionDetails;
    }

    public T Details
    {
        get;
        private set;
    }

    public DomainException()
    {
        // no argument constructor required for serialization
    }

    public DomainException(T exceptionDetails, Exception innerexception) : base(default(String), innerexception)
    {
        Details = exceptionDetails;
    }

    protected DomainException(SerializationInfo si, StreamingContext sc) : base(si, sc)
    {
        if (si == null)
            throw new NullReferenceException("si");

        Details = (T)si.GetValue("exceptionDetails", typeof(T));
    }
}

This exception class acts as a container for another simple property-only class that includes the exception details:

[Serializable]
public class MyExceptionDetails
{
    public MyExceptionDetails(int failureCode)
    {
        FailureCode = failureCode;
    }

    public int FailureCode
    {
        get;
        private set;
    }
}

A specific error details class like this one would be created where, in the past, an entire exception would have been created. Only the [Serialization] attribute is required to make it serialize-able/compatible with DomainException<T>.  Throwing this exception would look like:

throw new DomainException<MyExceptionDetails>(new MyExceptionDetails(1001));

And because the exception is a concrete type it can be explicitly caught:

try
{
    // some code here
}
catch(DomainException<MyExceptionDetails> ex)
{
    // examine ex.Details
}

Alternative to Visual Studio Solution Folders

Visual Studio provides solution folders as a means to organize items in a hierarchical fashion, especially items that are not part of any one project. This works when existing items are added to solution folders, but not new items. This is because solution folders are virtual - the hierarchy of folders as they appear within the solution is not mirrored on disk. In fact, no folders are created on the physical disk and an item that is added to a solution folder is simply placed in the root solution directory.

To work around the solution folder deficiency I simply created a new project to hold ancillary items. Specifically: 

• Create a new C# class library project called "Accessories".  
• Remove all classes from the new project.
• From the solution configuration manager, un-check the project's build flag for all configurations.
• Add folders and items to the project, setting each item's build action property to "None".

Viola, now the hierarchy of the folders and items within the project will be mirrored on disk.