Stephan Zahariev's Blog

Migrating ASP.NET MVC 1.0 to MVC 2.0: Real World Scenario

2010-04-28T13:47:00.001+03:00

As most of you have noticed, ASP.NET MVC v.2.0 has been released last month. The new version introduces lots of cool features, so most of the existing MVC 1 applications will be upgraded to the new release. In the current post I will try to share my experience with migrating existing ASP.NET MVC 1.0 application to MVC 2.0.

Scenario: There is an existing ASP.NET MVC 1.0 web application build on top of .NET Framework 3.5, jQuery and Visual Studio 2008. The goal is to migrate the application to MVC 2, while keeping the other libraries and tools (.NET 3.5, VS2008, etc).

Identifying the breaking changes

The first step from the process would be to check the ASP.NET MVC 2.0 breaking changes. Naturally after carefully evaluating each item in the breaking changes list, I have find out that the following will be a problem: “JsonResult now responds only to HTTP POST requests”. The problem was caused by a jQuery plug-in that uses only HTTP GET. So the solution was to:

Replace the plug-in by someone more configurable that can use HTTP POST. You should do it in case that your application exposes sensitive information and is vulnerable to the attack described here.
Explicitly allow HTTP GET on JsonResults. You can do it by using the JsonRequestBehavior.AllowGet

[AcceptVerbs(HttpVerbs.Get)]
public JsonResult GetData()
{
    //Some other code
    return Json(data, JsonRequestBehavior.AllowGet);
}

I was lucky that the information exposed in my application was not sensitive, so I decided to use the JsonRequestBehavior.AllowGet.

Migrating the solution to build against ASP.NET MVC 2.0

Being confident that I have resolved all breaking changes I had to think about migrating the solution to use the new version of MVC 2.0. If you don’t want to do everything by hand, you should use this tool. The tool is build by one of the Microsoft employees and works GREAT. However, I have noticed a few gotchas. The tool insists to backup your project before the conversion. This seems redundant, because usually the source code stays in code repository and if somehow the conversion produces a mess, everything could be restored. After all I had to wait a few extra minutes for the backup (more than 1GB in my case). The second issue that I have found is the update of the jQuery files. The tool updates the jQuery and Microsoft AJAX libraries. Since the web site was relying on several other 3rd party jQuery plug-ins I didn’t want the jQuery upgrade. So I had to manually remove the update. Despite of the above, the tool is absolutely FANTASTIC an you will need it.

Running the application

Up to now, everything went pretty well and the solution compiled without problems. So I was ready to run it. After hitting F5 I was stunned. The application started to close and open pages by itself! With the help of a few unit tests and Goolge I was able to identify the source of the problem. It turns out that there is another undocumented breaking change: a value from TempData dictionary will be removed after the request in which it is read! You can read more here. The fix was relatively easy and soon everything was working as usual.

Bottom line

The migration process from ASP.NET MVC 1.0 to MVC 2.0 is relatively easy and you should do it. The new features are awesome.

ASP.NET MVC Best Practices – Routes to Ignore

2010-03-26T01:54:00.003+02:00

Routing is a key feature from the ASP.NET MVC. If you are not familiar with this concept, take a look here, because I’m not going to explain it. I’m going to share some routes that you should ignore in your configuration.
So maybe you are wondering, why should I ignore some routes? And the simple answer is: Because most of the time, different search engines/crawlers/bots will try to index your site and they will request for specific files. For example, Google will try to access the following file: http://yoursite.com/robots.txt. The robots.txt file is used to to give instructions about your site to web robots. More info can be found here. Up to now, everything sounds great, but if your site does not have robots.txt file, the ASP.NET MVC framework will raise an exception upon a request like the above. So your log, where the unhandled exceptions are reported can be flooded with messages like:
System.Web.HttpException : A public action method 'robots.txt' could not be found
on controller 'YourPage.Controllers.YourController'.
Obviously, there are two ways to handle this – provide a dummy robots.txt file or instruct the ASP.NET MVC framework to ignore such URL’s. I prefer the second approach. In this post I will try to summarize all the URL’s that should be ignored by the ASP.NET application. So, here they are:

~/robots.txt
~/sitemap
~/sitemap.gz
~/sitemap.xml
~/sitemap.xml.gz
~/google_sitemap.xml
~/google_sitemap.xml.gz
~/favicon.ico
~/mobile
~/iphone
~/apple-touch-icon.png
~/sitemap.txt
~/sitemap.rss

You need to ignore the routes above, ONLY if your site do not uses such a file. If you are not familiar with the sitemaps concept, here is a quick overview that answers to most of the questions.
Here is the code, taken from the Global.asax.cs file that should ignore some of the URL’s outlined above:

public static void RegisterRoutes(RouteCollection routes)
{
routes.IgnoreRoute("{resource}.axd/{*pathInfo}");

routes.IgnoreRoute("robots.txt");
routes.IgnoreRoute("sitemap");
routes.IgnoreRoute("sitemap.gz");
routes.IgnoreRoute("sitemap.xml");
routes.IgnoreRoute("sitemap.xml.gz");
routes.IgnoreRoute("google_sitemap.xml");
routes.IgnoreRoute("google_sitemap.xml.gz");
routes.IgnoreRoute("favicon.ico");

//Rest of the code is ommited
}

I will update the list above as soon as I find a new URL that should be ignored, so stay tuned. As well you are welcome to contribute to this list by posting a comment.

Some SharePoint 2010 Articles

2010-02-14T13:04:00.002+02:00

Few months ago I have installed the beta version of the SharePoint 2010. The product looks really cool, although some of the requirements like: “must have Windows 2008 64bit” are really annoying.

In the past few weeks I have published some articles in one new site dedicated to SharePoint technologies called SharepointMonitor.com. If somebody is interested, here is the list:

I have plans for several other articles, so stay tuned. I will update this post when something new comes out.

NHibernate: Display executed SQL at the bottom of an ASP.NET page (MVC or WebForms)

2010-01-17T01:06:00.002+02:00

Almost every modern application that uses relational database is build on top of ORM tool. The ORM tools can significantly decrease the efforts for interacting with the database, but unfortunately most of the developers are not familiar with the SQL code that is produced by the ORM tools. I’m not hiding that my favorite ORM is NHibernate. I’m using it since early beta, but from time to time even I’m surprised by the produced SQL. So, you need to monitor closely what SQL is executed. This blog post will describe one not widely used but very useful approach – the executed SQL will be displayed as a footer on any ASP.NET web page. It does not matter if the web page is using MVC or WebForms. Interested? Then continue reading.

Currently, there are several ways to display the SQL produced by NHibernate:

Database profiler tool – some of the modern RDBMS systems are coming with a profiler tool that can hook to the database engine and display the executed SQL. For example SQL Server comes with SQL Server Profiler. Unfortunately the SQL Server Express edition is missing this tool.
NHibernate Profiler – this is a third party tool that hooks to your application and monitors the NHibernate. Comes with cool WPF UI interface and lots of other features. Unfortunately at the time of writing you will need to purchase a license to use it. As well you will need to switch from your application to the tool after each action to monitor the produced SQL.
Log4Net – you can configure the log4net library to save the produced SQL to a file for example. Unfortunately reading long files with SQL statements is too boring. Sometimes it is tricky to figure out the SQL executed as a result of single action.
Display the executed SQL at the bottom of an ASP.NET web page. The idea is to have an IHttpModule module that will inject the SQL code at the end of each request. This approach saves time, because the SQL is visible instantaneously after the action is executed. Unfortunately this trick works only for web applications.

There are several ways to access the SQL that has been executed by NHibernate. I will use a NHibernate interceptor. The code for the interceptor looks like this:

public class NHibernateSQLMonitor : EmptyInterceptor
{
    public static void Init(Configuration config)
    {
        config.SetInterceptor(new NHibernateSQLMonitor());
    }

    private static StringBuilder mExecutedSQL = new StringBuilder();

    public static string ExecutedSQL
    {
        get
        {
            return mExecutedSQL.ToString();
        }
    }

    public static void ClearExecutedSQL()
    {
        mExecutedSQL = new StringBuilder();
    }

    public override NHibernate.SqlCommand.SqlString OnPrepareStatement(
        NHibernate.SqlCommand.SqlString sql)
    {
        mExecutedSQL.AppendLine(sql.ToString());

        return base.OnPrepareStatement(sql);
    }
}

The important method is OnPrepareStatement. This method is invoked every time NHibernate executes SQL statement. The sql parameter holds the SQL that will be executed.

Before using it, the interceptor must be registered. One possible way to do it is like this:

NHibernate.Cfg.Configuration cfg = new NHibernate.Cfg.Configuration();
NHibernateSQLMonitor.Init(cfg);

ISessionFactory sessionFactory = cfg.BuildSessionFactory();

Now everything is ready for our IHttpModule that will inject the executed SQL at the bottom of a web page. Here is the implementation:

public class NHibernateSQLMonitorModule : IHttpModule
{
    public void Init(HttpApplication context)
    {
        context.PostRequestHandlerExecute += new EventHandler(
  PostRequestHandlerExecute);
    }

    void PostRequestHandlerExecute(object sender, EventArgs e)
    {
        HttpContext httpContext = ((HttpApplication)sender).Context;
        HttpResponse response = httpContext.Response;

        if (response.StatusCode == 302)
        {
            //Browser performs redirect. Do nothing.
            //The executed SQL will be shown on the next page
        }
        else if (response.ContentType == "text/html")
        {
            response.Write("<hr>");
            response.Write("<b>SQL Executed by NHibernate</b>");
            response.Write("<br>");

            string executedSQL = NHibernateSQLMonitor.ExecutedSQL.Replace(
   "\n", "<br>");
            response.Write(executedSQL);

            NHibernateSQLMonitor.ClearExecutedSQL();
        }
    }

    public void Dispose()
    {
        //Not required
    }
}

Now there is one final step that should be done – the HttpModule needs to be registered in the config file of the application. To do it open the web.config file, find the <httpModules> section (<modules> if you are running IIS7) and place the following line there:

<add name="NHibernateSQLMonitor"
 type="Data.NHibernateSQLMonitorModule, Data"/>

Data is the assembly name where the NHibernateSQLMonitorModule class is defined. Now everything is ready and if there are no compilation errors you should see something like this:

After adding new record, the page will look like this:

So, what about improvements:

Instead of using NHibernate interceptor for monitoring the executed SQL it is possible to display the output from the Log4Net. The log4net output contains the SQL parameter values which may be a great benefit. As you may have noticed, the values in the screenshot above are replaced by the ? character.

The current version uses static variable to hold the executed SQL. If there is more than one user interacting with the application, this will lead to some synchronization problems. For example, you can extend the NHibernateSQLMonitor class to use the ASP.NET Session store. This way the user will see only the SQL code executed as a result of his actions.

P.S. This post was inspired long time ago after reading the Steve Sanderson’s Pro ASP.NET MVC Framework. He has a similar application that connects to Linq to SQL and displays the executed SQL.

eBay Architecture

2009-12-09T16:54:00.001+02:00

Recently I was curious to find out how big sites like eBay are build and running. After a quick research I was able to find some quite interesting articles.

I was surprised to find out that eBay doesn't use transactions. As well they do not use foreign keys in their database (the same thing as the SharePoint_AdminContent database). It looks like this is the latest trend ;-)

If you are curious like me, take a look at the following:

Randy Shoup on eBay's Architectural Principles
eBay Architectural Strategies, Patterns, and Forces
Software Engineering Radio: eBay's Architecture Principles with Randy Shoup
Scalability Best Practices: Lessons from eBay
Dan Pritchett on Architecture at eBay
You Scaled Your What?
High Scalability: eBay Architecture

jQuery FullCalendar and ASP.NET MVC

2009-08-24T23:48:00.003+03:00

Recently I had to integrate jQuery FullCalendar into ASP.NET MVC application. Up to now I was not able to find such an example, so I will try to provide one. I assume that you are familiar with ASP.NET MVC, jQuery and FullCalendar component, so I’m not going to introduce each technology.

I will use the default ASP.NET MVC Web Application template to create new project. Please, note that the jQuery is included by default, so there is no need to download and reference it.

The first step would be to place FullCalendar JavaScript and style-sheet files inside project’s Scripts and Content directory as shown in the picture.

Then we need to reference the files we just added to our project. I will do it by placing the following lines of code inside the <head> section of Site.Master file:

<head runat="server">
   <title><asp:ContentPlaceHolder ID="TitleContent" runat="server" /></title>
   <link href="../../Content/Site.css" rel="stylesheet" type="text/css" />
   <link href="../../Content/fullcalendar.css" rel="stylesheet" type="text/css" />
   <script src="../../Scripts/jquery-1.3.2.js" type="text/javascript"></script>
   <script src="../../Scripts/fullcalendar.js" type="text/javascript"></script>
</head>

Now we are ready to use the calendar routines inside our views. How to do it? Just create a div tag and render the calendar’s HTML code inside. Here is how to modify Home’s Index.aspx:

<asp:Content ID="indexContent" ContentPlaceHolderID="MainContent" runat="server">

   <script type="text/javascript">
       $(document).ready(function() {
           $('#calendar').fullCalendar({
               events: "/Home/CalendarData"
           });
       });  
   </script>

   <div id="calendar">
   </div>
</asp:Content>

The code above will render the calendar’s HTML inside a div with id=”calendar”. The calendar data will be delivered by invoking the following URL: /Home/CalendarData. This corresponds to CalendarData method from the Home controller. This controller is supposed to return the data in Json format. Here is a sample implementation:

    [HandleError]
   public class HomeController : Controller
   {
       public ActionResult CalendarData()
       {
           IList<CalendarDTO> tasksList = new List<CalendarDTO>();

           tasksList.Add(new CalendarDTO
           {
               id = 1,
               title = "Google search",
               start = ToUnixTimespan(DateTime.Now),
               end = ToUnixTimespan(DateTime.Now.AddHours(4)),
               url = "www.google.com"
           });
           tasksList.Add(new CalendarDTO
           {
               id = 1,
               title = "Bing search",
               start = ToUnixTimespan(DateTime.Now.AddDays(1)),
               end = ToUnixTimespan(DateTime.Now.AddDays(1).AddHours(4)),
               url = "www.bing.com"
           });

           return Json(tasksList);
       }

       private long ToUnixTimespan(DateTime date)
       {
           TimeSpan tspan = date.ToUniversalTime().Subtract(
    new DateTime(1970, 1, 1, 0, 0, 0));

           return (long)Math.Truncate(tspan.TotalSeconds);
       }

       public ActionResult Index()
       {
           return View();
       }

       public ActionResult About()
       {
           return View();
       }
   }

The code above creates two calendar entries called “Google search” and “Bing search”. Everything should be pretty simple, except the stuff around ToUnixTimespan routine.

There is well known problem with serialization of dates in Json format. There is no strict standard, so there are several approaches to this problem. For example, take a look here. The implementation adopted by Microsoft was not recognized by FullCalendar, so I had to introduce the ToUnixTimespan routine. Basically, this routine returns the seconds after 1/1/1970.

Because of the above, you should notice that the start and end dates are represented as int:

    public class CalendarDTO
   {
       public int id { get; set; }
       public string title { get; set; }
       public long start { get; set; }
       public long end { get; set; }
       public string url { get; set; }
   }

If you have done everything correct, the final result will be

Enjoy!

Anti-development methodologies

2007-06-21T16:36:00.000+03:00

Software industry is growing constantly and development methodologies are following the same trend. Internet is full with shiny articles describing each methodology and all the benefit. Most of the people I'm talking are bored when they hear stuff like this. If you are like them, don't miss this article. A friend of mine just send me the link. Also, check out the comments. All of them are about dummy development practices. Some of the titles are cynical, but there is a lot of truth in each line.
Enjoy!

BrainStorm Rocks

2007-04-27T10:57:00.000+03:00

Yesterday was a great day. My team called BrainStorm won the local round of Microsoft's Imagine Cup 2007 competition in the area of software design.
Our project with the code name Bookvar will compete in the finals this year in Seoul, Korea. This was possible, because of the remarkable team I'm part of. Right now I can't find appropriate words to describe them, but as one of the jury said: "I had the chance to watch these young motivated and smart people". It's a great pleasure to work with each one of them.
Although we managed to win the local round, there is still much work to be done. Lucky for us, the finals are after 3 months.
Check out this post from Ruslan Trifonov if you want to see some pictures from the event. Ruslan was member of the jury.

Grand Prix Racing Online

2007-03-13T11:55:00.000+02:00

Last week I was chatting with an old friend of mine - Vladimir Alexandrov. I was surprised to find out that he is driving the Grand Prix Racing Online site. Basically this is a F1 on-line simulator. If you are a F1 fan you will appreciate what Vladimir and his team is doing. The site is completely free and guess what - no advertisements are present.
There are some interesting issues about the implementation and hosting, but because of security I can not share them.
Good luck Vlado, and continue with the great job.

NHibernate and triggers

2007-03-12T17:34:00.000+02:00

Recently I was asked from a college of mine, about the following exception in application based on SQL Server 2005 and NHibernate :

[NHibernate.AdoNet.TooManyRowsAffectedException] {"Unexpected row count: 2; expected: 1"} NHibernate.AdoNet.TooManyRowsAffectedException

After a quick investigation the problem was obvious - the exception raises when a row from a table is updated and the row has fired a trigger which executes an update statement.
I have faced with this problem since early beta of NHibernate (0.8.4 as far as I can remember). Till now the only solution I have found is to use SET NOCOUNT to stop the message that shows the number of rows affected by a Transact-SQL statement from being returned as part of the results.
Just place SET NOCOUNT ON before the UPDATE statement inside the trigger and then after the UPDATE statement place SET NOCOUNT OFF. This will solve the problem.

ASP.NET Web Part editor in pop-up dialog

2007-03-12T17:29:00.000+02:00

Last hour I was wondering how to host a ASP.NET Web Part editor control inside a pop-up dialog. I was lucky to find this implementation. It is based on some tricks, but resolves the problem.

First Bulgarian .NET Book is here

2006-12-20T16:46:00.000+02:00

Second volume from the first Bulgarian book about .NET is out. You can download free copy from here. Printed version will be available after a month.

Windows Forms data binding and notification of changed properties

2006-11-17T11:46:00.000+02:00

Yesterday a college of mine asked me a question: "How to refresh TextBox control in Windows forms application binded to a property of object, when the property change its value?". The solution is very simple, but according to me not very well documented (at least in books I read).

Solution: Provide event named PropertyNameChanged, where PropertyName is the name of property binded to the TextBox. During the binding process, the text box inspects binded object for event with the mentioned signature and binds to it. The sample code looks like this:

public class Customer

    private string mName;

    public string Name

        get { return mName; }

set

            mName = value;

            if (NameChanged != null)

                NameChanged(this, EventArgs.Empty);

    public event EventHandler NameChanged;

Customer cust = new Customer();

textBox1.DataBindings.Add("Text", cust, "Name");

Another possible solution is to remove and add binding of the text box to the object. You can write general routine to loop through each control for a given form and remove-add binding. This is nasty solution of course.

Cool conference in Bulgaria

2006-09-11T20:21:00.000+03:00

DevReach - one of the most exiting conferences about Microsoft technologies in Bulgaria is coming. According to the program, some of the world leading industry speakers will present quite interesting lectures. Hope we will meet there.

Google toolbar 4

2006-09-11T20:01:00.000+03:00

This morning I was amazed! A few minutes after turning on my PC a small balloon appeared. The message in the balloon was something about Google toolbar update. Who cares, was my first reaction. But when the update was finished I was surprised to learn that LONG WAITING FEATURES LIKE ACCESSING BOOKMARKS FROM ANY COMPUTER are already present. Believe me - this is like a elixir of life for me. I'm using one PC at work, one PC at home and one laptop. Managing bookmarks in situation like this is almost nightmare. In the past there were some other decision of this problem, but none of them was so elegant as Google toolbar. BTW, Microsoft are developing the same service, but as far as I know this feature is still in beta.
The other features I like are SpellCheck and Translate. If you are inspired, install the toolbar from here for WinXP and IE. Enjoy!

ADO.NET Entity Framework ContextObject and ASP.NET application

2006-09-04T09:15:00.000+03:00

Last month Microsoft published first CTP of ADO.NET Entity Framework. For more info - look here. Although Microsoft is still planning integration between Entity Framework and ASP.NET, I want to show you how to handle ADO.NET Entity Framewrok context object in ASP.NET application. This post is inspired from NHibernate session management.
The idea is simple: Context object will be stored in ASP.NET session and after the request HTTPModule will free it. I'm using ASP.NET session instead of ASP.NET context to further extend this example to support session per application transaction strategy. If you are wondering what is "session per application transaction" take a look at NHibernate session management. Here is the code of the implementation:

public class ContextManager<T>where T: ObjectContext, new()

    public static T ObjectContext

get

            T resultContext;

            if (HttpContext.Current.Items.Contains(

                ContextConst.CONTEXT_OBJECT_SESSION_KEY))

                resultContext =

                    (T)HttpContext.Current.Items[

                        ContextConst.CONTEXT_OBJECT_SESSION_KEY];

            else

                resultContext = new T();

                HttpContext.Current.Items[ContextConst.CONTEXT_OBJECT_SESSION_KEY] =

                    resultContext;

            return resultContext;

    public static bool isActiveObjectContext

get

            return HttpContext.Current.Items.Contains(

                ContextConst.CONTEXT_OBJECT_SESSION_KEY);

public class ContextHttpModule : IHttpModule

    #region IHttpModule Members

    public void Init(HttpApplication context)

        context.EndRequest += new EventHandler(OnEndRequest);

    public void Dispose()

        //Intentionaly not implemented

    public void OnEndRequest(Object sender, EventArgs e)

        if (HttpContext.Current.Items.Contains(ContextConst.CONTEXT_OBJECT_SESSION_KEY))

            ObjectContext contextObject =

                (ObjectContext)HttpContext.Current.Items[

                    ContextConst.CONTEXT_OBJECT_SESSION_KEY];

            contextObject.Dispose();

    #endregion

internal class ContextConst

    public const string CONTEXT_OBJECT_SESSION_KEY =

        "ContextConst.CONTEXT_OBJECT_SESSION_KEY";

Now, lets look how the code above can be used. First lets make some corrections to web.config file:

<connectionStrings>

  <add name="Duwamish7Model.Duwamish7"

       connectionString="metadata=C:\Temp\ContextObject\Model;

          mapping=C:\Temp\ContextObject\Model;

          provider=System.Data.SqlClient;

          provider connection string=&quot;Data Source=ns-server;

          Initial Catalog=Duwamish7;Integrated Security=True&quot;"

    providerName="System.Data.Mapping" />

</connectionStrings>

<httpModules>

  <add name="ContextObjectManager"

       type="ObjectContextManager.ContextHttpModule, ObjectContextManager"/>

</httpModules>

Having all of the above done, we can write somethig like this:

Duwamish7 db = ContextManager<Duwamish7>.ObjectContext;

Query<Authors> query = db.GetQuery<Authors>(

    "SELECT VALUE a FROM Authors AS a WHERE a.PKId > 144");

Repeater1.DataSource = query;

Repeater1.DataBind();

P.S. The code above is not tested for concurrency problems.

WCF and WS Eventing

2006-09-03T16:48:00.000+03:00

WCF (aka Indigo) is the latest communication technology from Microsoft (still in beta). Although there are a lot of goodies in this technology, something really useful for me is missing - implementation of WS Eventing. If you are wondering why - this article gives the answer. Fortunately there is already non-Microsoft implementation. Read this article if you are interested.

Finally it is true!

2006-08-24T10:27:00.000+03:00

Yesterday, PFC Levski Sofia did it. The team had succeeded to qualify for the UEFA champions league group stage for the first time in Bulgarian football history. Although I'm fan of the CSKA Sofia the success of Levski makes me happy. More info - here.

Microsoft SQL Server and Ad-Hock queries optimization

2006-08-23T18:36:00.000+03:00

Yesterday I was surprised to learn that Microsoft SQL Server 2005 makes behind the scene parameterization of ad-hock queries. For example, lets have the following query:

SELECT ID, Name FROM Article WHERE Price > 100

SQL Server 2005 will replace constant literal values (100 in this example) by variables and this way query plan will be reuses for all queries similar to

SELECT ID, Name FROM Article WHERE Price > @p1

Not all of the ad-hock queries can be parameterized. If you are interested read this article.

Attribute based transaction management and NHibernate

2006-08-23T17:31:00.000+03:00

In my previous post about NHibernate session management I presented a session per request strategy for session management in ASP.NET application. I will extend this example to show you how to declaratively manage transactions via attributes. Take a look at the GenericDAO.Save method:

public void Save(T entity)

    bool manageTransaction = ! orm.isActiveTransaction;

    if (manageTransaction)

        orm.BeginTransaction(IsolationLevel.ReadCommitted);

try

        orm.session.SaveOrUpdate(entity);

        if (manageTransaction)

            orm.CommitTransaction();

    }catch(Exception e)

        if (manageTransaction)

            orm.RollbackTransaction();

        throw e;

Almost 90% of the code is for transaction management. It will be nice if we can write something like:

[Transaction(IsolationLevel.ReadCommitted)]

public void Save(T entity)

    mOrm.session.SaveOrUpdate(entity);

    mOrm.session.Flush();

To achieve this, I will use Spring.NET to intercept method call of the DAO. Before method marked with TransactionAttribute is executed a transaction will be started and then committed after method is executed. Although .NET provides solution for intercepting method calls via ContextBoundObject, I will use Spring.NET because this way my DAO object must implement just one interface instead of inheriting from ContextBoundObject.
Lets begin with the implementation. First I will declare the attribute used to mark transaction sensitive methods:

[AttributeUsage(AttributeTargets.Method)]

public class TransactionAttribute : Attribute

    private IsolationLevel mIsolationLevel;

    public TransactionAttribute()

        mIsolationLevel = IsolationLevel.ReadCommitted;

    public TransactionAttribute(IsolationLevel aIsolation)

        mIsolationLevel = aIsolation;

    public IsolationLevel isolationLevel

        get { return mIsolationLevel; }

        set { mIsolationLevel = value; }

Now I will decrare the class responsible for starting and committing transactions. There are a lot of specific issues about method interception and Spring.NET so if you are not familiar with them read the documentation about Spring.NET. The class looks like:

public class TransactionAroundAdvice : IMethodInterceptor

    OrmManager mOrm;

    public TransactionAroundAdvice()

        mOrm = OrmManagerFactory.GetInstance();

    private TransactionAttribute GetTransactionAttribute(MethodInfo aMethod)

        object[] attributes = aMethod.GetCustomAttributes(

            typeof(TransactionAttribute), true);

        if (attributes.Length != 1)

            //TODO: Throw more meaningfull exception and message

            throw new Exception("Transaction Attribute is missing");

        return (TransactionAttribute)attributes[0];

    public object Invoke(IMethodInvocation invocation)

        bool manageTransaction = ! mOrm.isActiveTransaction;

        if (manageTransaction)

            TransactionAttribute transactionAttribute =

                GetTransactionAttribute(invocation.Method);

            mOrm.BeginTransaction(transactionAttribute.isolationLevel);

        //TODO: Find way to compare isolation level from TransactionAttribute and

        //current active transaction if manageTransaction == false

        object returnValue;

try

            returnValue = invocation.Proceed();

            if (manageTransaction)

                mOrm.CommitTransaction();

        }catch(Exception e)

            if (manageTransaction)

                mOrm.RollbackTransaction();

            if ((e is TargetInvocationException) && (e.InnerException != null))

                //Because when AOP is enables, method is invoked via reflection

                //TargetInvocationException is thrown. This exception is most likely

                //to confuse application developer, so it is removed

                throw e.InnerException;

            else

                throw e;

        return returnValue;

When a method marked with TransactionAttribute is intercepted then Invoke method from TransactionAroundAdvice class is called. Now lets see how our DAO class looks like when transaction management code is removed:

public class AttributeEnabledDAO<T>: IDAO<T>

    public AttributeEnabledDAO()

        mOrm = OrmManagerFactory.GetInstance();

    private OrmManager mOrm;

    #region IDAO Members

    public T Load(object id)

        return (T) mOrm.session.Load(typeof(T), id);

    [Transaction(IsolationLevel.ReadCommitted)]

    public void Save(T entity)

        mOrm.session.SaveOrUpdate(entity);

        mOrm.session.Flush();

    #endregion

Much more simple. Example of using AttributeEnabledDAO is given bellow:

ProxyFactory articleDAOFactory = new ProxyFactory(

    new AttributeEnabledDAO<Article>());

articleDAOFactory.AddAdvisor(new DefaultPointcutAdvisor(

    new AttributeMatchMethodPointcut(typeof(TransactionAttribute)),

    new TransactionAroundAdvice()));

IDAO<Article> articleDao;

lock (synchObject)

    articleDao = (IDAO<Article>) articleDAOFactory.GetProxy();

Article article = articleDao.Load(1);

article.Name = "Paracetamol";

article.SalePrice = rnd.NextDouble();

articleDao.Save(article);

ADO.NET vNext CTP is out

2006-08-16T10:48:00.000+03:00

Microsoft has just released ADO.NET vNext CTP. This CTP includes the preview build of the long waiting ADO.NET Entity Framework. This Entity Framework is Microsoft's ORM tool. If you are wondering what is ORM tool take a look at my post about
NHibernate session management.
Presentation with features of Entity Data Model can be found here. If somehow Link* technologies looks messy for you, taka a look at this presentation.

Querying Microsoft SQL Server trace files

2006-08-06T09:06:00.000+03:00

Have you ever wanted to be able to query saved SQL Server profiler trace files (*.trc)? Yes! Then you need to use fn_trace_gettable.

Interview with Anders Hejlsberg

2006-08-05T11:48:00.000+03:00

Take a look at this interview with Anders Hejlsberg. It's not about a future technology. It's about himself. Watching this video you will find answers to questions like: Why Anders left Borland and joined Microsoft? What is the secret of his success? Anders talks for the time working on projects like: Turbo Pascal, Delphi, C#. Enjoy!

Mocking static methods

2006-08-05T10:19:00.000+03:00

Yes, the title is correct. Mocking static methods in .NET is possible with TypeMock library. There is no need to redesign you classes just to test them. In fact there is no need to write additional code to your classes. This is possible because TypeMock uses profiler's API to handle execution of methods and creation of classes. For a list of features look here.
Using TypeMock is pretty easy. In my previous post about NHibernate session management I'm using a static method to create appropriate session manager (take a look at GenericDAO constructor). The test of the constructor may be something like:

[TestFixture]

public class GenericDAOTest

    [SetUp]

    public void InitTest()

        MockManager.Init();

        //ormManagerMock = MockManager.Mock(typeof(PerRequestWebManager));

        ormManagerFactory = MockManager.Mock(typeof(OrmManagerFactory));

    //private Mock ormManagerMock;

    private Mock ormManagerFactory;

    private GenericDAO<Article> articleDAO;

    [TearDown]

    public void FinalizeTest()

        MockManager.ClearAll();

    [Test]

    public void TestDAOConstructor()

        ormManagerFactory.ExpectAndReturn("GetInstance",

            new PerRequestWebManager());

        articleDAO = new GenericDAO<Article>();

        MockManager.Verify();

Although TypeMock is powerful and easy to use library, there are some disadvantages. For example in the free version, when the test is started for first time, a popup dialog appears and invites you to buy the product. Also not all of the .NET types can be mocked (for more information take a look at this thread).

NHibernate Session Management

2006-07-25T21:08:00.001+03:00

I'm big fan of NHibernate since early beta. Unfortunately this tool is not very popular among .NET developers, so I'm planning to blog a lot about this framework. For those of you who don't know, NHibernate is ORM (Object-Relational-Mapping) tool. Basically ORM tools are used to "persist" C# objects into RDBMS system. More detail explanation of ORM tools is available here.
NHibernate handles loading/saving objects to RDBMS via it's own implementation of session object. Until now, I'm familiar with 3 ways for handling NHibernate sessions:

create session per each database operation - session is created before and destroyed after each operation with database. This is very ineffective way, because some features like lazy loading are unavailable. Also including several database operations in transaction is impossible.
create session per each request - session is created before and destroyed after each HTTP request. This is my favorite strategy, because it is effective and very easy for implementation. I'll give example for possible implementation in this post.
create session per each application transaction - session is created before and destroyed after each application transaction. Example for application transaction is adding invoice with invoice items.

So, lets begin with the implementation of session per request strategy. If you are reading this post you are most likely a developer and probably as most developers prefer to read source code instead of boring documentation. So I'll explain all of the classes in brief and will post full source code. First I'll declare a base abstract class for all of the session strategies.

public abstract class OrmManager

    public OrmManager()

    static NHibernate.Cfg.Configuration mConfiguration = null;

    static volatile ISessionFactory mSessionFactory = null;

    static object mFactorySyncObject = new object();

    public abstract ISession session

        get;

    public IDbConnection sessionConnection

        get { return session.Connection; }

    public ISession NewSession()

        ISessionFactory sessionFactory = GetSessionFactory();

        return sessionFactory.OpenSession();

    public abstract bool isActiveSession

        get;

    public abstract void closeSession();

    private const string HBM_ASSEMBLY_NAME_SECTION = "HbmAssemblyName";

    private ISessionFactory GetSessionFactory()

        if (mSessionFactory == null)

            lock (mFactorySyncObject)

                if (mSessionFactory == null)

                    string hbmAssemblyName =

                        ConfigurationManager.AppSettings[HBM_ASSEMBLY_NAME_SECTION];

                    mConfiguration = new NHibernate.Cfg.Configuration();

                    mConfiguration.AddAssembly(hbmAssemblyName);

                    mSessionFactory = mConfiguration.BuildSessionFactory();

        return mSessionFactory;

    public void BeginTransaction(IsolationLevel isolation)

        session.BeginTransaction(isolation);

    public void CommitTransaction()

        session.Transaction.Commit();

    public void RollbackTransaction()

        session.Transaction.Rollback();

    public bool isActiveTransaction

get

            return ((session.Transaction != null) &&

                (!session.Transaction.WasCommitted) &&

                (!session.Transaction.WasRolledBack));

Next, I'll implement the class responsible for session per request management. The following class implements IHttpModule so, we can handle end of the request and close any open sessions.

public class PerRequestWebManager : OrmManager, IHttpModule

    public PerRequestWebManager()

    private const string NHIBERNATE_SESSION_KEY = "NHIBERNATE_SESSION";

    public override ISession session

get

            ISession resultSession;

            if (HttpContext.Current.Items.Contains(NHIBERNATE_SESSION_KEY))

                resultSession =

                    (ISession)HttpContext.Current.Items[NHIBERNATE_SESSION_KEY];

            else

                resultSession = NewSession();

                HttpContext.Current.Items[NHIBERNATE_SESSION_KEY] = resultSession;

            return resultSession;

    public override bool isActiveSession

        get { return HttpContext.Current.Items.Contains(NHIBERNATE_SESSION_KEY); }

    public override void closeSession()

        if (isActiveSession == true)

            if (isActiveTransaction == true)

                RollbackTransaction();

            session.Close();

            HttpContext.Current.Items.Remove(NHIBERNATE_SESSION_KEY);

    #region IHttpModule

    public void Init(HttpApplication context)

        context.EndRequest += new EventHandler(OnEndRequest);

    public void Dispose()

        //Intentionaly not implemented

    public void OnEndRequest(Object sender, EventArgs e)

        closeSession();

    #endregion

Now I'll make one generic DAO class with the help of previos PerRequestWebManager.

public class GenericDAO<T>

    public GenericDAO()

        orm = OrmManagerFactory.GetInstance();

    private OrmManager orm;

    public T Load(object id)

        return (T) orm.session.Load(typeof(T), id);

    public void Save(T entity)

        bool manageTransaction = ! orm.isActiveTransaction;

        if (manageTransaction)

            orm.BeginTransaction(IsolationLevel.ReadCommitted);

try

            orm.session.SaveOrUpdate(entity);

            if (manageTransaction)

                orm.CommitTransaction();

        }catch(Exception e)

            if (manageTransaction)

                orm.RollbackTransaction();

            throw e;

public class OrmManagerFactory

    private OrmManagerFactory()

    private const string ORM_MANAGER_TYPE_SECTION = "OrmManager";

    public static OrmManager GetInstance()

        string ormType = ConfigurationManager.AppSettings[ORM_MANAGER_TYPE_SECTION];

        if (ormType == typeof(PerRequestWebManager).Name)

            return new PerRequestWebManager();

        else

            throw new Exception("Unknown ORM manager type");

And here's how all of the above can be used in ASP.NET application:

GenericDAO<Article> articleDao = new GenericDAO<Article>();

Article article = articleDao.Load(1);

article.Name = "Paracetamol";

article.SalePrice = rnd.NextDouble();

articleDao.Save(article);

If you are interested I can mail you full source of this post. Enjoy!