Nikhil Singhal's Blog

Basics of Search Engine Optimization

Nikhil Singhal — Sun, 08 Mar 2009 08:06:00 GMT

Google has a great document that talks in extreme clear terms what site owners should do to enable better indexing of their content by majore search engines. This document can be found at http://googlewebmastercentral.blogspot.com/2008/11/googles-seo-starter-guide.html...(read more)

How to: Identify Blocked SQL Processes Quickly

Nikhil Singhal — Sun, 13 Jul 2008 23:29:00 GMT

There was a great article in Visual Studio magazine (June 2008) by Ian Stirk in which he talks in detail about how to improve application performance by creating a utility that tells you which processes are being blocked.

You can read the article at http://visualstudiomagazine.com/features/article.aspx?editorialsid=2490. The two sql sprocs that you will need to create are:

The database utility dba_BlockTracer extracts data about the running processes by inspecting the system view sys.sysprocesses. This view then queries the underlying system table sysprocesses.

CREATE PROC [dbo].[dba_BlockTracer]

/*--------------------------------------------------

Purpose: Shows details of the root blocking process, together with details of any blocked processed

----------------------------------------------------

Parameters: None.

Revision History:

19/07/2007 Ian_Stirk@yahoo.com Initial version

Example Usage:

1. exec YourServerName.master.dbo.dba_BlockTracer

--------------------------------------------------*/

BEGIN

-- Do not lock anything, and do not get held up by any locks.

SET TRANSACTION ISOLATION LEVEL READ

UNCOMMITTED

-- If there are blocked processes...

IF EXISTS(SELECT 1 FROM sys.sysprocesses WHERE

blocked != 0)

BEGIN

-- Identify the root-blocking spid(s)

SELECT distinct t1.spid AS [Root blocking spids]

, t1.[loginame] AS [Owner]

, master.dbo.dba_GetSQLForSpid(t1.spid) AS

'SQL Text'

, t1.[cpu]

, t1.[physical_io]

, DatabaseName = DB_NAME(t1.[dbid])

, t1.[program_name]

, t1.[hostname]

, t1.[status]

, t1.[cmd]

, t1.[blocked]

, t1.[ecid]

FROM sys.sysprocesses t1, sys.sysprocesses t2

WHERE t1.spid = t2.blocked

AND t1.ecid = t2.ecid

AND t1.blocked = 0

ORDER BY t1.spid, t1.ecid

-- Identify the spids being blocked.

SELECT t2.spid AS 'Blocked spid'

, t2.blocked AS 'Blocked By'

, t2.[loginame] AS [Owner]

, master.dbo.dba_GetSQLForSpid(t2.spid) AS

'SQL Text'

, t2.[cpu]

, t2.[physical_io]

, DatabaseName = DB_NAME(t2.[dbid])

, t2.[program_name]

, t2.[hostname]

, t2.[status]

, t2.[cmd]

, t2.ecid

FROM sys.sysprocesses t1, sys.sysprocesses t2

WHERE t1.spid = t2.blocked

AND t1.ecid = t2.ecid

ORDER BY t2.blocked, t2.spid, t2.ecid

END

ELSE -- No blocked processes.

PRINT 'No processes blocked.'

END

Make a call to the database function dba_GetSQLForSpid to get the underlying SQL, which can show you why performance is slow. The function accepts one parameter: the SQL Server process ID (@spid) of a running process.

CREATE Function [dbo].[dba_GetSQLForSpid]

(

@spid SMALLINT

)

RETURNS NVARCHAR(4000)

/*-------------------------------------------------

Purpose: Returns the SQL text for a given spid.

---------------------------------------------------

Parameters: @spid - SQL Server process ID.

Returns: @SqlText - SQL text for a given spid.

Revision History:

01/12/2006 Ian_Stirk@yahoo.com Initial version

Example Usage:

SELECT dbo.dba_GetSQLForSpid(51)

SELECT dbo.dba_GetSQLForSpid(spid) AS [SQL text]

, * FROM sys.sysprocesses WITH (NOLOCK)

--------------------------------------------------*/

BEGIN

DECLARE @SqlHandle BINARY(20)

DECLARE @SqlText NVARCHAR(4000)

-- Get sql_handle for the given spid.

SELECT @SqlHandle = sql_handle

FROM sys.sysprocesses WITH (nolock) WHERE

spid = @spid

-- Get the SQL text for the given sql_handle.

SELECT @SqlText = [text] FROM

sys.dm_exec_sql_text(@SqlHandle)

RETURN @SqlText

END

Running a Windows Service from command line

Nikhil Singhal — Thu, 10 Jul 2008 01:44:00 GMT

One of the common problems that we face in designing a Windows service is the ease of debugging it. I have followed a pattern to solve this problem where I can run a service from either command line or as a service.

The steps below outline the changes you would need to make to enable this in your service. Also included are code snippets that show the point.

1. Modify your Service.

a. Add a RunConsole() method that performs a OnStart(), waits for user to quit and a OnStop() call.

b. Add a RunService() method to Run the process as a normal service.

/// <summary>

/// Runs the service as a console app.

/// </summary>

internal static void RunConsole()

{

using (MyService consoleService = new MyService())

{

consoleService.OnStart(null);

// Wait for the user to quit the program.

Console.WriteLine("Press \'q\' to quit");

while (Console.Read() != 'q') ;

consoleService.OnStop();

}

/// <summary>

/// Runs normally

/// </summary>

internal static void RunService()

{

System.ServiceProcess.ServiceBase.Run(new MyService());

}

2. Modify your startup program.

a. Update your Main(string[] args) to accept arguments, and if the first argument is a /c, call RunConsole() (above), else call RunService().

/// <summary>

/// The main entry point for the service...

/// It will run the executable as a service

/// unless /c is specified in which case it will run as a console app.

/// </summary>

static int Main(string[] args)

{

int retCode = 0;

if (args.Length > 1)

{

PrintUsage();

retCode = 1;

}

else if (args.Length == 1)

{

if ((args[0] != null) && (args[0].Length >= 2))

{

switch (args[0][0])

{

case '-':

case '/':

switch (args[0][1])

{

case 'c':

case 'C':

MyService.RunConsole();

break;

default:

PrintUsage();

retCode = 1;

break;

}

break;

default:

PrintUsage();

retCode = 1;

break;

}

else

{

PrintUsage();

retCode = 1;

}

else

{

MyService.RunService();

}

return retCode;

}

That's it.

How to AutoIncrement version with each build using Team Foundation Server build (with a little help from AssemblyInfoTask)

Nikhil Singhal — Fri, 13 Jun 2008 06:40:00 GMT

One common requirement with any decent sized multi-version product is to automatically update the version numbers of the binaries on a regular basis. This is generally achieved by updating the AssemblyInfo.cs (or other language equivalent ) files.

There are a couple of ways to do this:

1. Assign one developer to remember to increment the numbers on a daily basis. Yes, don’t get mad at me. I know this sucks. I just listed it here as an option.

2. Use the nightly build (or equivalent tool) to do this automatically.

If you use Team Foundation Server to manage your daily (or nightly builds), you can use one custom task developed by the nice people at Microsoft. The task is aptly named AssemblyInfoTask. This task used to be hosted at www.gotdotnet.com but since it closed people often to go www.codeplex.com to search for it. Well, it now resides at code.msdn.com. The direct link to the project is http://code.msdn.microsoft.com/AssemblyInfoTaskvers.

Once you download it and follow the instructions and set it up, you will run into the next problem. You will see an error similar to this:

Error emitting 'System.Reflection.AssemblyVersionAttribute' attribute -- 'The version specified '1.0.08102.1' is invalid

This is an interesting problem. See, we have had version numbering since ages. During those early days, when disk space was a premium, a good developer never wasted space. So, they ended up using Int16 for the build number (the 3^rd set of numbers). This caused a huge problem with the arrival of 2007.

The solution for this problem is described here: http://blogs.msdn.com/msbuild/archive/2007/01/03/fixing-invalid-version-number-problems-with-the-assemblyinfotask.aspx

Ok. Now you have got this build to happen on your dev box. How do you set this up on your build server?

Simple. Note: The project documentation says to add it to the *.csproj file (or *.vbproj file). DON’T DO THAT if you are going to use the build server for nightly builds. Let’s open your TFSBuild.proj file (again, if you used the defaults, it will be under your root dir of the project/TeamBuildTypes/Release (or Debug). Let’s check out this file and open it.

At the very top of the file (after comments and start), you will see a line like:

ADD the following line after this:

But this is not all. If you run the build now, you will see errors of File not accessible. The reason, as you would have guessed, is that AssemblyInfo.* files would be under source control as read only and could not be saved. The solution now is to check out these files before the compile gets executed and check them back in after compile is done.

Here is the sample config on how to do this. Note that in my example below, both C# and VB AssemblyInfo files will get covered.

Step 1: In the <PropertyGroup> section, add the following:

<TF>"$(TeamBuildRefPath)\..\tf.exe"</TF>

<AssemblyInfoSpec>AssemblyInfo.*</AssemblyInfoSpec>

</PropertyGroup>

The first line defines an alias for tf.exe to checkout/checkin the files.

The second line defines a file pattern.

Step 2: We need to add <targets>. You will need to add these lines towards the end of the project (after the <ItemGroup> sections.

</CreateItem>

<Exec WorkingDirectory="$(SolutionRoot)"

Command="$(TF) checkout /recursive $(AssemblyInfoSpec)"/>

</Target>

Hopefully, it’s apparent what’s happening here. The target tells the build system to execute this task after getting all the source code. The CreateItem creates an array of files that meet the file spec we defined earlier. The xec checks these out.

Now that we have added the checkout, let’s check the files in after AssemblyInfoTask has done the edits. Let’s add the following lines:

<Exec WorkingDirectory="$(SolutionRoot)"

Command="$(TF) checkin /comment:"Auto-Build: Version Update" /noprompt /override:"Auto-Build: Version Update" /recursive "/>

</Target>

Good so far. What happens if you build broke. Ah-ha, you see where I am going.

<Exec WorkingDirectory="$(SolutionRoot)"

Command="$(TF) undo /noprompt /recursive $(AssemblyInfoSpec)"/>

</Target>

Now we have covered this properly. Let’s check this file in and fire a build. If you have enabled check-in mails, you should get a mail that will tell about AssemblyInfo files getting updated.

Happy building.

Visual Studio Setup/deployment projects and Team Foundation Server

Nikhil Singhal — Thu, 12 Jun 2008 09:04:00 GMT

Team Foundation servers use MSBuild to build our projects. MSBuild does not support the Visual Studio Setup/Deployment projects natively. In many of today's applications it's a must to have msi based installs created via Visual Studio.
Till someone actually takes the time to build a custom task library to handle vdproj files in MSBuild, we will have to use a hack to make our vdproj files get built on MSBuild. The hack is simple. After the compilation of the main solution is done, we invoke the Visual studio command line to build the vdproj project and copy the msi and setup.exe to the appropriate output folder.
The following simple steps will get you going in the correct direction.

Find your TFSBuild.proj file in TFS source control for your project. Generally, if you had selected all defaults when creating a build definition, it would reside under TFSProject/TeamBuildTypes/{Build/Release} folder.
Make sure your vdproj project is part of the main solution. Or else, you could create a standalone solution with the same name as the vdproj project. Either way,
Edit the TFSBuild.proj file and add the following XML snippet (generally after the <ItemGroup> node).
Replace the folder and files in the snippet with your project specific paths and names. Also, in the example below, I have shown the build type to be {Release/Any CPU}. You would replace this with your build specific settings.
Check this file back in and fire a build. You are good to go!!!

Handling global web service unhandled exceptions

Nikhil Singhal — Wed, 11 Jun 2008 19:47:00 GMT

One of the most tiresome (but important) things when developing web services is handling un-handled exceptions. A good design principle forces you to catch and cast relevant exceptions raised by your web methods into more meaningful SOAP exceptions. But exceptions will occur.

It is quite tedious to wrap each web method in a try/catch loop. This dictates a need for a common framework to handle unhandled web services exceptions. In this blog post, I will guide you through a step by step process for building one.

Summary:

Extend SoapExtension class and override the ProcessMessage() method.
In ProcessMessage(), add special handler code for message stage of SoapMessageStage.AfterSerialize.
Modify the web service's web.config file and add a <soapExtensionTypes> node to <webServices> section.

Details:

Ok, now let's dig deeper into the code.

Step 1:

The first step in the whole process is to extend the SoapExtension class override the ProcessMessage() method.

using System.IO;
using System.Web.Services;
using System.Web.Services.Protocols;
using System.Xml;
using System;

public class TelspaceSoapExtension : SoapExtension
{
private Stream originalStream;
private Stream updatedStream;

    public override object GetInitializer(Type serviceType)
    {
        return null;
    }

    public override object GetInitializer
    (LogicalMethodInfo methodInfo, SoapExtensionAttribute attribute)
    {
        return null;
    }

    public override void Initialize(object initializer)
    {
    }

    public override Stream ChainStream(Stream stream)
    {
        originalStream = stream;
        updatedStream = new MemoryStream();
        return updatedStream;
    }

    private void Transfer(Stream inStream, Stream outStream)
    {
        StreamReader sr = new StreamReader(inStream);
        StreamWriter sw = new StreamWriter(outStream);
        sw.Write(sr.ReadToEnd());
        sw.Flush();
    }

    public override void ProcessMessage(SoapMessage soapMessage)
    {
        switch (soapMessage.Stage)
        {
            case SoapMessageStage.BeforeDeserialize:
                Transfer(originalStream, updatedStream);
                updatedStream.Position = 0;
                break;

            case SoapMessageStage.AfterSerialize:
                if ((soapMessage.Exception != null))
                {
                    ExceptionProcessor processor = new ExceptionProcessor();

string details;

// handle our exception, and get the SOAP <detail> string
details = processor.HandleWebServiceException(soapMessage);

                    // read the entire SOAP message stream into a string
                    updatedStream.Position = 0;
                    TextReader tr = new StreamReader(updatedStream);

                    // insert our exception details into the string
                    string s = tr.ReadToEnd();
                    s = s.Replace("<detail />", details);

                    // overwrite the stream with our modified string
                    updatedStream = new MemoryStream();
                    TextWriter tw = new StreamWriter(updatedStream);
                    tw.Write(s);
                    tw.Flush();
                }

                updatedStream.Position = 0;
                Transfer(updatedStream, originalStream);
                break;
        }
    }

}

Step 2:

The next step is to handle the exception and get out meaningful details from the exception. For this purpose, let's dig deeper into the ExceptionProcessor class. This class has one public method: HandleWebServiceException(System.Web.Services.Protocols.SoapMessage sm). This method is called from our ProcessMessage()case SoapMessageStage.AfterSerialize.

Here is the code:

using System;
using System.Collections.Generic;
using System.Text;
using System.Xml;
using System.Web.Services.Protocols;
using System.Runtime.InteropServices;
using System.Diagnostics;
using System.Reflection;
using SASMTPLib;
using System.IO;

class ExceptionProcessor
{
private string currentExceptionDetails;
private string currentExceptionTypeName;

private const string RootExceptionName = "System.Web.HttpUnhandledException";
private const string RootWsExceptionName = "System.Web.Services.Protocols.SoapException";

    public string HandleWebServiceException(System.Web.Services.Protocols.SoapMessage sm)
    {
        HandleException(sm.Exception);
        XmlDocument doc = new XmlDocument();
        XmlNode detailNode = doc.CreateNode(XmlNodeType.Element, SoapException.DetailElementName.Name, SoapException.DetailElementName.Namespace);
        XmlNode typeNode = doc.CreateNode(XmlNodeType.Element, "ExceptionType", SoapException.DetailElementName.Namespace);
        typeNode.InnerText = currentExceptionTypeName;
        detailNode.AppendChild(typeNode);

        XmlNode messageNode = doc.CreateNode(XmlNodeType.Element, "ExceptionMessage", SoapException.DetailElementName.Namespace);
        messageNode.InnerText = sm.Exception.Message;
        detailNode.AppendChild(messageNode);

        XmlNode infoNode = doc.CreateNode(XmlNodeType.Element, "ExceptionInfo", SoapException.DetailElementName.Namespace);
        infoNode.InnerText = currentExceptionDetails;
        detailNode.AppendChild(infoNode);
        return detailNode.OuterXml.ToString();
    }

    private void HandleException(Exception ex)
    {
        try
        {
            currentExceptionDetails = NormalizeException(ex);
            currentExceptionTypeName = ex.GetType().FullName;

            // ignore root exceptions
            if (((currentExceptionTypeName == RootExceptionName) || (currentExceptionTypeName == RootWsExceptionName)))
            {
                if (ex.InnerException != null)
                {
                    currentExceptionTypeName = ex.InnerException.GetType().FullName;
                }
            }
        }
        catch (Exception e)
        {
            currentExceptionDetails = string.Format("Error \'{0}\' while generating exception string", e.Message );
        }

ExceptionToEmail();
}

    private string NormalizeException(Exception ex)
    {
        StringBuilder sb = new StringBuilder();

        if (ex.InnerException != null)
        {
            if (((ex.GetType().ToString() == RootExceptionName) || (ex.GetType().ToString() == RootWsExceptionName)))
            {
                return NormalizeException(ex.InnerException);
            }
            else
            {
                sb.Append(NormalizeException(ex.InnerException));
                sb.Append(Environment.NewLine);
                sb.Append("(Outer Exception)");
                sb.Append(Environment.NewLine);
            }
        }

        // get exception-specific information
        sb.Append("Exception Type:\t\t");
        try
        {
            sb.Append(ex.GetType().FullName);
        }
        catch (Exception e)
        {
            sb.Append(e.Message);
        }
        sb.Append(Environment.NewLine);

        sb.Append("Exception Message:\t\t");
        try
        {
            sb.Append(ex.Message);
        }
        catch (Exception e)
        {
            sb.Append(e.Message);
        }
        sb.Append(Environment.NewLine);

        sb.Append("Exception Target Method:\t");
        try
        {
            sb.Append(ex.TargetSite.Name);
        }
        catch (Exception e)
        {
            sb.Append(e.Message);
        }
        sb.Append(Environment.NewLine);

        sb.Append("Stack Trace:\t\t");
        sb.Append(Environment.NewLine);
        try
        {
            sb.Append(ex.StackTrace);
        }
        catch (Exception e)
        {
            sb.Append(e.Message);
        }
        sb.Append(Environment.NewLine);

return sb.ToString();
}

    private void ExceptionToEmail()
    {
        // Send e-mail code
    }

}

Step 3:

The final step is to modify web.config of the web services to include the following section (under <system.web> node)

Bulk Insert into SQL from C# App

Nikhil Singhal — Wed, 11 Jun 2008 19:43:00 GMT

One of the common problems I have seen is to bulk upload data to a SQL Server database. If you have the flexibility to directly run your code in SQL, you have a ton of options. But let's say that you have to massage the data before you throw it in to the database, then you have to really know your SQL (well to do it in SQL).

Let's say you have to read data from an RSS feed, parse it and then load it into SQL. Let's assume further that this feed updates every 2 hours. It would be a trivial task to write a C# app that reads and parses the feed. One crude way to upload this data would be to do a single row insert for each data element. This would be terribly inefficient. The other option would be to use .Net framework's SqlBulkCopy class.

The basic template would be something like

private void WriteToDatabase()
{
    // get your connection string
    string connString = "";
    // connect to SQL
    using (SqlConnection connection = 
            new SqlConnection(connString))
    {
        // make sure to enable triggers
        // more on triggers in next post
        SqlBulkCopy bulkCopy = 
            new SqlBulkCopy
            (
            connection, 
            SqlBulkCopyOptions.TableLock | 
            SqlBulkCopyOptions.FireTriggers | 
            SqlBulkCopyOptions.UseInternalTransaction,
            null
            );

        // set the destination table name
        bulkCopy.DestinationTableName = this.tableName;
        connection.Open();

        // write the data in the "dataTable"
        bulkCopy.WriteToServer(dataTable);
        connection.Close();
    }
    // reset
    this.dataTable.Clear();
    this.recordCount = 0;
}

The above code snippet shows you the API usage. But before you actually do that, you need to follow a couple of steps to setup your data table.

First, let's look at a simple record structure (as reflected in C# class):

using System;
using System.Data;
using System.Configuration;

/// <summary>
/// Summary description for MyRecord
/// </summary>
public class MyRecord
{
    public int TestInt;
    public string TestString;

	public MyRecord()
	{
	}

    public MyRecord(int myInt, string myString)
    {
        this.TestInt = myInt;
        this.TestString = myString;
    }
}

Now, let's start dissecting the class that we will use to upload the data:

using System;
using System.Data;
using System.Collections.Generic;

using System.Data.SqlClient;
using System.Configuration;
using System.IO;

namespace SqlExamples.FileLoader
{
/// <summary>
/// Summary description for BulkUploadToSql
/// </summary>
public class BulkUploadToSql
{
    private List<MyRecord> internalStore;

    protected string tableName;
    protected DataTable dataTable = new DataTable();
    protected int recordCount;
    protected int commitBatchSize;

Note that we have an internal List data structure as well as the DataTable. This is redundant and you can avoid using the internalStore if your application does not need to massage the data before it's sent to SQL.

I then define 2 private constructors. The reason is that we want to use the factory pattern to return our object to the caller.

    private BulkUploadToSql(
        string tableName, 
        int commitBatchSize)
    {
        internalStore = new List<MyRecord>();

        this.tableName = tableName;
        this.dataTable = new DataTable(tableName);
        this.recordCount = 0;
        this.commitBatchSize = commitBatchSize;

        // add columns to this data table
        InitializeStructures();
    }

    private BulkUploadToSql() : 
        this("MyTableName", 1000) {}

Note that we set the commit batch size. This is a very important factor that needs to be fine tuned for your database. What this defines is the number of records that we would send in one shot to the database.

The next step is to Initialize the data table with columns that reflect the actual table structure.

private void InitializeStructures()
{
    this.dataTable.Columns.Add("TI", typeof(Int32));
    this.dataTable.Columns.Add("TS", typeof(string));
}

I then provided a factory method to load data into my internal structure from a data source. In the example code below, I use a Stream, but this can be any data source from where you wish to populate your data.

public static BulkUploadToSql Load(Stream dataSource)
{
    // create a new object to return
    BulkUploadToSql o = new BulkUploadToSql();
    
    // replace the code below
    // with your custom logic 
    for (int cnt = 0; cnt < 10000; cnt++)
    {
        MyRecord rec = 
            new MyRecord
            (
            cnt, 
            string.Format("string{0}", cnt)
            );
        o.internalStore.Add(rec);
    }

    return o;
}

This would make sure that our class is properly initialized and loaded with data. Once the caller has a valid object, they can now "Flush" the data as shown below:

public void Flush()
{
    // transfer data to the datatable
    foreach (MyRecord rec in this.internalStore)
    {
        this.PopulateDataTable(rec);
        if (this.recordCount >= this.commitBatchSize)
            this.WriteToDatabase();
    }
    // write remaining records to the DB
    if (this.recordCount > 0)
        this.WriteToDatabase();
}

private void PopulateDataTable(MyRecord record)
{
    DataRow row;
    // populate the values
    // using your custom logic
    row = this.dataTable.NewRow();

    row[0] = record.TestInt;
    row[1] = record.TestString;

    // add it to the base for final addition to the DB
    this.dataTable.Rows.Add(row);
    this.recordCount++;
}

In the example above, the call to Flush() actually massages the data (and at the same time loads it into the actual data table). As I mentioned before, you can actually skip this step if your application does not require massaging.

As a example of an app that uses this class:

using System;
using System.Collections.Generic;
using System.Text;
using SqlExamples.FileLoader;
using System.IO;

namespace DemoApp
{
class Program
{
    static void Main(string[] args)
    {
        using (Stream s = 
            new StreamReader(@"C:\TestData.txt"))
        {
            BulkUploadToSql myData = 
                BulkUploadToSql.Load(s);
            myData.Flush();
        }
    }
}
}

As always, this is JUST demo code to explain a concept. This is NOT production quality code and please make sure to follow the coding guidelines in your team.

Happy coding....