Using Entity Framework Code First with a dynamic connection string

Ok, I've managed to find a solution myself and I'm quite happy with it.

First, I'll lay out the structure of the application.

I do have a DbContext called AppServerSettingsContext which is defined in the source code like this:

/// <summary>
///     Initializes a new instance of the <see cref="AppServerSettingsDataContext"/>.
/// </summary>
public AppServerSettingsDataContext()
    : base("AppServerSettingsDataContext")
{ }

This context has 2 different entities (one for a member and one for all the settings for that specific member).

In order for the migration to be able to execute, I need to have a ConnectionString in the applications configuration file, just like the way we used.

Then, I have another context named AppServerDataContext. This has 2 constructors as showed below:

/// <summary>
///     Initializes a new instance of the <see cref="AppServerDataContext"/>.
/// </summary>
public AppServerDataContext() : 
    this(ConfigurationManager.ConnectionStrings["AppServer"].ConnectionString) 
{ }

/// <summary>
///     Initializes a new instance of the <see cref="AppServerDataContext"/>.
/// </summary>
/// <param name="connectionString">The full connection string which is used to connect to the database.</param>
public AppServerDataContext(string connectionString)
    : base(connectionString) { }

You'll see in the code that I can either specify the connection string or I load the connection string for the application configuration file.

You'll see why this is important later on.

I have the DbContext which points to a connection string inside a configuration file. It's not required to have this, but I'm used to work that way. So that connection string will ONLY be used when calling the add-migration command. This is because that command requires a database to check the current state of the database and add the correct migration.

Now, I'm working inside one single project with 2 context files, so the NuGet Package Manager Console needs a way to identify this.

Therefore, the following commands can be used:

1. Enable migrations for a specific context:

   • `PM> Enable-migrations -ContextTypeName: -MigrationsDirectory:

This is a command which I need to execute twice, once for every context.

Then, in the Seed method of my AppServerSettings context, I'll write the following code:

/// <summary>
///     Runs after upgrading to the latest migration to allow seed data to be updated.
/// </summary>
/// <param name="context">Context to be used for updating seed data.</param>
protected override void Seed(AppServerSettingsDataContext context)
{
    // Creates the Member and assign all the settings which are required for the application to function.
    context.Members.AddOrUpdate(x => x.Name, new Member("Povlo")
    {
        Settings = new List<MemberSettings>
        {
            new MemberSettings("DatabaseConnectionString", "Removed for Security Reasons"),
        }
    });

    // Make sure that for every member, the database is created by using the "MigrateDatabaseToLatestVersion" migration.
    foreach (var setting in context.Members.Select(member => member.Settings.FirstOrDefault(x => x.Key == "DatabaseConnectionString")))
    {
        using (var appServerContext = new AppServerDataContext(setting.Value))
        {
            var de = new MigrateDatabaseToLatestVersion<AppServerDataContext, AppServer.Configuration>();
            de.InitializeDatabase(appServerContext);

            appServerContext.Database.Initialize(true);
        }
    }
}

What I do here basically is first creating a member with a given connection string (there can be multiple).

Then, in the same method, I do have a foreach loop, which will create a context based on the connection string in the database. Then for this context, the database is being upgraded to the latest version.

The advantage of this is that I'm using a code-first approach, and all the databases are always the latest version.

A disadvantage of this is that all models needs to be exactely the same.