Dinesh's Blog ::::::: Being Compiled :::::::: SQL Server 2005

Analysis Services Browser Views Add-In

Tue, 03 Jun 2008 03:45:05 GMT

Yossi Elkayam has done a nice add-in to the BIDS that allows you to save views created for SSAS and run them as new requests. Seems this is very useful if you are currently working on a SSAS project.

Read more and download from here.

Nuggets from my SSIS experience

Wed, 26 Mar 2008 00:42:26 GMT

I have been doing lot of development on SSIS for last two, three months and I made this note on what I faced during the project. It may useful for your future SSIS development, or brush up your SSIS knowledge. Read and see.

Click here to read the article.

Anyway, original name for the article was not this :), thanks for Gogula, naming it with nice title.

Data Comparison with Data Dude

Tue, 11 Mar 2008 00:59:48 GMT

My latest article, Data Comparison with Data Dude has been published at www.sql-server-performance.com. This talks about the way of comparing the data between two data sources and scripting the changes.

Read and see, click here to visit the site.

Faced another issue with SSIS

Mon, 10 Mar 2008 05:22:09 GMT

Had to spend few minutes on this, to resolve the issue but found no solution. The problem was really weird. It was with variable window;

Look at the variable window. All buttons are disabled, disallowing creating or modifying variables created. I made a search to see whether anyone has experienced this before but could not find anything related to this. However, closing and reopening the packed resolved the problem :).

Taking advantages of Instant File Initialization

Tue, 26 Feb 2008 01:20:13 GMT

Have you ever tried this and seen the speed of file allocation process? Though I had heard about it, I had not tried (I might have but I cannot remember :)). The default file allocation of SQL Server is zero initialization that takes long time when you try to create a database with a large sized data file or when you try to restore a fairly large database that is not exist. These two processes (not only these two, other processes like adding a new file...) can be speeded up by enabling Instant Initialization. You can enable the Instant Initialization by giving proper permission to the server that runs the SQL Server instance. Let me show you the simple test I did;

My SQL Server service runs with a account called SQLServerAccount that has just few user rights. It is not in the administrators group. I ran the code below and it took 13 seconds to create the "Test" database.

CREATE DATABASE [Test] ON PRIMARY
( NAME = N'Test', FILENAME = N'C:\Program Files\Microsoft SQL Server\MSSQL.1\MSSQL\DATA\Test.mdf' , SIZE = 200 , FILEGROWTH = 1000 )
LOG ON
( NAME = N'Test_log', FILENAME = N'C:\Program Files\Microsoft SQL Server\MSSQL.1\MSSQL\DATA\Test_log.ldf' , SIZE = 1024KB , FILEGROWTH = 10%)
GO

I dropped the database and added the SQLServerAccount user account to "Perform volume maintenance tasks" policy. it is located under Administrative tasks -> Local security settings -> User right assignment. After restarting the SQL Server service, I ran the script again. Now SQL Server avoids zero initialization. It took only 1 second to create the database. This is a just a database with 200MB, so, think about a database with 10GB, or think about a restore process of 10GB database. Hope you can see the advantage.

This is not enabled by default because of security reason. Enabling this allows SQL Server administrator to see the recently deleted files that are not supposed to be seen by the SQL Server administrator.

This feature only available with SQL Server 2005 and it needs to be run with Windows 2003, XP or later. If the service account is already the local administrator, you already have the advantage. Note that this is not applicable to the log file.

11th SS SLUG meeting - Scheduled to 21st

Tue, 19 Feb 2008 18:57:12 GMT

Since the 20th is a Poyaday that is a mercantile holiday in Sri Lanka, we decided to have SS SLUG meeting on 21st, Thursday. Fiqri, ASP.NET MVP is going to join with us this time, showing his knowledge on SQL Server mobile edition. I am sure that Fiqri is going to deliver very valuable info for us because we have not discussed this topic before.

Gogula will be doing a session too; "Inmon vs. Kimbal - The great debate". This is a very interesting topic, and I am sure that people involve with BI know about this debate. Let's see what he is going to deliver us.

And I will be doing a session too; I will be showing another greatest redgate tool, SQL Multi Script.

Hope we have added enough things to the meeting. As usual, we have arranged few things for give-away items, including T-shirts, USB drives, Mugs, Pens. Come and join with us. Listen and grab both knowledge and prizes.

For more info, visit SQL Server Universe .Com

Another article at sql-server-performance

Fri, 08 Feb 2008 01:13:22 GMT

Another note of mine talks about "Schema Comparison with Data Dude" has been published at www.sql-server-performance.com. This shows the way of doing the schema comparison and how Data Dude addresses the changes.

Read it from here and let me know whether it is good enough for readers.

Two new articles published

Wed, 23 Jan 2008 11:04:19 GMT

Two of my articles have been published; one at http://sql-server-performance.com and http://sqlserveruniverse.com.

Test Data Generation with Data Dude
This article explains the usage of Data Dude for test data generation. It is published at http://sqlserveruniverse.com.

SSIS: Usage of Checkpoint File
I wrote this few months back and it was in the queue. I made this note based on the experience I had when checkpoint files are used with SSIS.

Read both and see whether they are useful.

SQL Server 2005 Best Practices Analyzer (January 2008)

Sat, 19 Jan 2008 01:47:13 GMT

The latest version is available with new stuff; with new rules, enhanced facilities for documentation, more support for SSAS. I did not test anything yet but you ca start testing, using it.

Download the BPA from here.

9th Consecutive SS SLUG meeting - with Christmas Eve

Wed, 19 Dec 2007 03:47:27 GMT

SQL Server Sri Lanka User Group meeting is happening today (19th) and yes, it is the last meeting for this year. Gogula will be doing a session regarding SQL Server 2005 Service Broker that we really wanted cover with previous meetings and Dinesh Asanka will be doing a session about one of the red-gate products; SQL Data Compare.

We decided to give out lots of prizes this time, including mugs, t-shirts, USB pen drives, so pay attention and grab them all. All these give-away items are sponsored by red-gate, Microsoft Sri Lanka, Virtusa-eCollege and SQL Server Universe.Com.

New Article at SQL Server Universe

Fri, 23 Nov 2007 07:27:13 GMT

I published a new article regarding log shipping with SQL Server 2005. I have added details as much as possible and the way of implementation.

Read and see whether it is worthwhile.

Data Dude Presentation

Fri, 23 Nov 2007 07:22:32 GMT

I have uploaded the presentation I did at SS SLUG meeting. You can download the presentation from here.


Doing my presentation - part of the audience	Prize grabbers

The one Dinesh Asanka did will be uploading by Monday. See some pictures of the meeting here.

Black box in data type property

Mon, 22 Oct 2007 04:55:54 GMT

I found this with one of my Script Component. See the image;

If I click on it, the entire window goes off. Im trying to find a solution for this because this has blocked editing the type of the output. If anyone has seen this before and has found a solution, please let me know.

Data Mining presentation at SS SLUG meeting

Fri, 19 Oct 2007 05:46:23 GMT

The 7th SQL Server User Group meeting was held at Microsoft Sri Lanka, on last Wednesday. Gogula started the first session, delivered valuable info about new policy management in SQL Server 2008. Next was our special segment, demonstrating one of red-gate products. Dinesh Asanka did the session, exploring the value of SQLCompare. I did the last one, I talked about Data Mining with SQL Server 2005. It was an introductory session for Data Mining, talked about data mining algorithms and algorithms introduced by Microsoft for SQL Server.

We gave out few great prizes, including SQLCompare licensed copy that is worth more than US$ 500, 1GB USB pen drives, mugs and t-shirts. Prizes were sponsored by red-gate and Microsoft Sri Lanka.

Presentations related to these sessions will be published at www.sqlserveruniverse.com by end of week.

Lesson from a cartoon: SQL Injection

Fri, 19 Oct 2007 05:38:11 GMT

One of my friends had sent me this link, nice cartoon about SQL injection :)

http://xkcd.com/327/

Another article at sql-server-performance.com

Wed, 10 Oct 2007 04:12:08 GMT

The second note of Index Related DMVs and DMFs series has been published at www.sql-server-performance.com. This article shows the usage of sys.dm_db_index_usage_stats dynamic management view. Read the article and see.

Two new articles at sqlserveruniverse.com

Tue, 09 Oct 2007 04:13:40 GMT

We have added two new articles at www.sqlserveruniverse.com, one is written by Gogula and other from me. Here are the links;

Populating Dimension Tables with the SCD Task by Gogula G. Aryalingam
ETLing dimensions through SQL Server 2000 DTS, has more or less been quite a a hassle, what with the lengthy ActiveX scripts and lookups that have to be performed. Enter SQL Server 2005 Integration Services: The Slowly Changing Dimension task has made it much simpler. As simple as a few clicks and minutes. This article talks about how a slowly changing dimension table can be populated using the Slowly Changing Dimension transformation task, which comes with Integration Services.

Reporting Services Data Regions: Table Data Region by Dinesh Priyankara
Organizing data with a combination of different data visualizations is a rather complicated yet important task in report authoring. SQL Server Reporting Services 2005 includes more than three data regions that allow us to organize data easily. This article talks about one of the data regions; Table Data Region.

Read them and see how worth they are; and dont forget to rate and comment.

sys.sysprocesses shows as "runnable" but sys.dm_exec_requests shows as "running"

Mon, 08 Oct 2007 12:59:54 GMT

The sys.dm_exec_requests is the replacement (or new implementation) of sys.sysprocesses system table. These two allow us to see the queries currently run under the instance of SQL Server. I noticed that the these two objects give different status for same execution; sys.sysprocesses give "runnable" while sys.dm_exec_requests give "running". You can simply test this one by running a batch and monitor via these two objects. I googled and tried to find a explanation for this but nothing found.

Any comment on this?

Add regions to your SQL codes

Tue, 02 Oct 2007 03:14:28 GMT

www.ssmstoolspack.com offers set of add-ins for SSMS that gives some great features. Since this is still in beta, not all functions are properly worked. One thing I really like is the feature that allows us to add regions in the code just like Visual Studio.

Since the region is added as a comment, saved scripts can be open with any SSMS and can execute without any problem. If the tool pack is not installed, it will be just a commented line without collapse/expand facility.

Download the toolspack and see. Since it is still in beta, better not to install in production environment.

Property QuotedIdentifierStatus is not available for UserDefinedFunction

Mon, 24 Sep 2007 06:23:48 GMT

I faced for this error today when I was trying to modify a SQL 2000 table-valued function using SQL 2005 management studio. You may have already seen/faced, yes it is a bug, not a problem with security or anything but it has been fixed with SP1; if you face, make sure you install the latest service pack.

Here is the link for it: https://connect.microsoft.com/SQLServer/feedback/ViewFeedback.aspx?FeedbackID=126099

Index related DMVs and DMFs

Thu, 20 Sep 2007 03:29:03 GMT

Another articles of mine is published at www.sql-server-performance.com. This explores output of sys.dm_db_index_physical_stats function. I have added examples as much as possible; to show the actual usage of the function.

Read the article: http://www.sql-server-performance.com/articles/per/index_dmv_dmf_p1.aspx

Another article at SQL Server Central

Tue, 21 Aug 2007 10:09:45 GMT

This note was made regarding .NET assemblies and SSIS. It is a short, simple note that explains the way of accessing .NET assemblies in SSIS and problems we might be facing. Read this and see whether it is useful.

Query shortcut key for SELECT * FROM

Wed, 08 Aug 2007 01:16:43 GMT

I rarely use query shortcut keys. The reason behind for that is, I have not added any new query shortcut keys :), only default keys are set. Recently there was a blog post about query shortcut keys in the SqlAdvice and it opened by eyes for using query shortcut keys. One cool setting is, you can set a part of a statement for one shortcut key. Then highlighting a word in the query editor and hitting the shortcut key will complete the statement defined in the shortcut key with the highlighted word and execute. Check this;

Open SQL Server Management Studio and open Options dialog box. Expand Environment node in the left pane and click on Keyboard. You can see default shortcut keys set for some built-in SPs. Select one of unused shortcut keys (eg. Ctrl+0) and write SELECT * FROM in the Stored Procedure column. Click OK to save the setting. Now open a new query window and write a table name (or drag a table from Object Explorer). Highlight it and hit Ctrl+0. Query window will execute a statement like SELECT * FROM {yourtable}. How easy? You can set different types of statement for other query shortcut keys. For more info, read this post.

IMPLICIT Transaction mode

Tue, 31 Jul 2007 06:59:53 GMT

One of the things I noticed at the last SQL Server Sri Lanka User Group meeting was, unawareness of IMPLICIT Transaction mode. Almost all were knew about AutoCommit and EXPLICIT modes but very few were aware of IMPLICIT mode. Are you aware of it? Then how often you use it?

The Difference between IMPLICT and EXPLICT is simple. EXPLICIT mode allows us to start the transaction wherever we want by using BEGIN TRAN statement. In IMPLICIT mode, the transaction automatically starts base on some T-SQL statements. Some of the statements are CREATE, INSERT, UPDATE, DELETE, ALTER TABLE. See the BOL for whole list. As EXPLICIT mode, either COMMIT or ROLLBACK is required to complete the transaction.

   1: USE Northwind   2: GO   3:     4: SET IMPLICIT_TRANSACTIONS ON   5: SET XACT_ABORT ON   6:     7: CREATE TABLE dbo.EmployeeAllowances   8:     (EmployeeId int REFERENCES dbo.Employees(EmployeeID) NOT NULL,   9:     MonthAndYear datetime NOT NULL,  10:     Allowance money NOT NULL)  11:    12:    13: INSERT INTO dbo.EmployeeAllowances  14:     (EmployeeId, MonthAndYear, Allowance)  15: VALUES  16:     (100, getdate(), 450.00)  17:    18:    19: COMMIT TRAN

The SET IMPLICIT_TRANSACTIONS ON tells the system to set the transaction mode to IMPLICIT. In above code, transaction starts with CREATE TABLE statement. Since I have turned XACT_ABORT on, transaction rolls back at any run-time error. If no run-time error, transaction is committed at COMMIT statement. In this case, everything is rolled back since INSERT statement raises an error.

SQL Server Sri Lanka User Group meeting - July 2007

Tue, 24 Jul 2007 01:00:39 GMT

It went okay. Chamindu's session was quite impressive. Though I have done some testing about SSIS custom components, I had no chance to implement a one for one of our applications. One thing I noted is, it is simple and very powerful.

In my presentation, one of my samples didn't work, because of the funny reason :(. I think that I have used a open-connection that had set to different isolation level, and tried to get a result-set related to another isolation level. I couldn't figure it out at that time, time might be reason because I had taken extra 15 minutes (according to Gogula :)). My presentation and samples (updated) are available for download, whoever wish to see Implementing Transaction with SQL Server 2005, visit this url.

SQL Server 2005 Best Practices Analyzer - July 2007

Mon, 09 Jul 2007 09:52:41 GMT

SQL Server 2005 BPA July one is available for download at here. If you want to know whether you have configured the server properly and need some recommendations, this is the best tool for it.

Does the SQL Server always resolve the objects names in SP?

Tue, 19 Jun 2007 04:01:07 GMT

You might already know this but thought to post it because it brushed up my knowledge and hope it brush up your knowledge too.

SQL Server starts its compiling process by checking the syntax of the code and then it checks the existence of the referenced objects. Unfortunately, this check does not always run (or it runs in different way). For example, you can create (or compile) a SP that has a reference to a table that is not exist but you cannot create a SP that has a reference to table that is exist but referred columns. Check the below code.

   1: create table TestTable100 (id int)   2: go   3:     4: -- this SP has reference for a table that is not exists (TestTable101)   5: -- but you can run this script and TestSP is successfully created   6: create proc TestSP   7: as   8: begin   9:    10:     select * from TestTable101  11: end  12: go  13:    14: -- this SP has a reference to the correct table but has referenced to the column   15: -- that is not exist.  16: -- SQL Server does not create this SP because it checks for column names  17: alter proc TestSP  18: as  19: begin  20:    21:     select name from TestTable100  22: end  23: go

Did it remind you something? Hope it does.

New white paper: Scale-Out Querying with Analysis Services

Mon, 18 Jun 2007 00:50:31 GMT

This is the latest article written about scaling out querying architecture in Analysis Services. Check this out. It is really worthwhile for reading.

Let the SSRS access SSIS package as data source

Tue, 12 Jun 2007 01:08:10 GMT

I had not tried this but the requirement came yesterday, see the possibility of setting SSIS package as a data source for SSRS reports. By default, it is not enabled, you will not see anything related to SSIS in the DataSource Type drop-down in the data source dialog box. Follow below set if you need to set SSIS packages as data source for your reports.

Open RSReportDesigner.config file through text editor. It is stored in the \Program Files\Microsoft Visual Studio 8\Common7\IDE\PrivateAssemblies folder.
Remove comment marks that have been set for SSIS extensions under data and designer element. Save and close it.
Open the rsreportserver.config file in the \Program Files\Microsoft SQL Server\MSSQL.3\Reporting Services\ReportServer folder.
Clear comment marks set for SSIS extension under data element.

Now you can access and process SSIS packages for reports. I may be writing complete procedure and publish but this is total enough for starting the experiment.

New article about Check Database Integrity Task

Fri, 25 May 2007 08:22:06 GMT

Wrote the first article to the SQL Server Universe, regarding SSIS. This is the first article of the series of articles. We will be adding more different types of articles soon but currently focusing on SSIS and SSAS. Read it and see whether it is worthwhile.

2nd SQL Server User Group Meeting in Sri Lanka is scheduled

Sat, 05 May 2007 03:31:54 GMT

The 2nd SSUG meeting is scheduled on 16th May, 2007. As usual, two sessions will be done; regarding replication and indexing. I'll be speaking about indexing, not about how to create them, but how to analyze and diagnosis. During my presentation, I'll be going covering some of new DMVs and dashboard facility too. Dinesh Asanka will be delivering the first session regarding replication. If you are in Sri Lanka, and interest, visit and see how useful these sessions are. For more info, visit http://www.sqlserveruniverse.com.

SQL Server Universe assisted to launch BI

Sat, 05 May 2007 03:23:07 GMT

I am glad to announce that SQL Server Universe did its first SQL training session for one of leading marketing companies in Sri Lanka. Gogula and myself ran four interactive sessions regarding BI, SSIS, SSAS and data analyzing with Office 2007 (including hands-on training). The goal of all these training sessions was implementing business intelligence with SQL Server 2005 product suite. The sessions went okay and seemed that the data mining part was the most interest for them.

We have some plans to launch SQL Server hands-on training for developers, will be announcing dates and places soon.

Indirect Configuration and multiple SSIS packages

Mon, 23 Apr 2007 04:34:01 GMT

One of the disadvantages with direct configuration (configuration location is hard-coded in the package) is, the location of the configuration file should be same as it was at the design time. For example, if you use the configuration file as D:\MyConfigFile\PackageConfiguration.dtsConfig at the design time, SSIS package expects the config file from the same path once it is deployed. One way to over some this issue is, disable the configuration from the package and compile, and assign the configuration file explicitly at run-time. Other way is, Indirect Configuration.

Indirect configuration can be simply implemented by setting the configuration location in the envoirnment variable. The configuration file can be deployed to any location. The SSIS package looks for the envoirnment variable for the location of the config file.

Applying same configuration file for multiple packages is possible. The header section of the configuration setting is totally ignored at the runtime (you can see the configuration section if you open the package with any text editor) and it tries to set all properties available in the config file to the package. You will see warnings if all properties cannot be applied to the package but can suppress all warnings by setting SuppressConfigurationWarnings true.

My SP does not use the index I added

Tue, 17 Apr 2007 06:43:55 GMT

I am sure that you have experienced this issue. You might have seen that the content of the SP works fine (within the expected time-frame), but not the SP. There are many reasons for this, but I recently noticed this issue in one of the SPs which I examined. Let me share the issue and fix with you, it might be helpful WITH your work.

Assume that the below table contains 100000 (from the year 1753) records and you try to get some records by using the SP GetTestData.

CREATE TABLE TestTable (ID int IDENTITY(1,1) PRIMARY KEY, Date datetime, String varchar(1000))
GO
CREATE INDEX IX_TextTable ON TestTable (Date)
GO

CREATE PROC GetTestData @number int
AS
BEGIN

DECLARE @NewDate datetime
SET @NewDate = DATEADD(month, @number, getDate())
SELECT ID, String FROM TestTable WHERE Date > @NewDate

END
GO

Note that we can add "DATEADD(month, @number, getDate())" to the WHERE condition without any problem but presume that the formula is much more complex than this and cannot be added to the "WHERE" condition directly.

Now let's try to get some result-sets. If I need to get records for last 10 months, I will be executing the below query;

EXEC GetTestData -10

If you have enabled "Include Actual Execution Plan" and "I/O", you may notice that SQL Server has not used the index which we had added and uses more logical reads than we expected (14340).

You may try the query itself by hardcoding the value;
SELECT ID, String FROM TestTable WHERE Date > DATEADD(month, -10, getDate())

Now you will see that it has used the index and has read few pages (938) only, and has executed within a lesser time-frame. So, where have we made the mistake? Why didn't SQL Server use the index when the SP was executed? The reason is simple. When the SP is compiled, initially it has no value for @number parameter and because of this it generates the plan for 30% of the records in the TestTable, and in this case it generates the plan that tells SQL Server to do a clustered index scan rather than scan the index we added. So, it never uses the index unless we force.

To overcome this behaviour, or to force using the index, we can use the OPTIMIZE FOR option with the statement. Note that you have to know about the exact requirment for the SP before applying the OPTIMIZE FOR with a value. For example, it would be better to do a clustered scan for a large result set than an index scan.

Here is the new SP;
CREATE PROC GetTestData @number int
AS
BEGIN

DECLARE @NewDate datetime
SET @NewDate = DATEADD(month, @number, getDate())
SELECT ID, String FROM TestTable WHERE Date > @NewDate OPTION (OPTIMIZE FOR (@NewDate = '01/01/2007'))

END

Now we can see that;

EXEC GetTestData -10

is executed by using the index we added and with the time we expected and less page reads. So, if you face this issue, measure it, understand the requirement, and use OPTIMIZE FOR.

Fix for the SSIS Script Task error....

Tue, 10 Apr 2007 10:49:37 GMT

This error put me into a big trouble when I was doing the SSIS presentation. I couldnt figure out the reason for the problem but finally found out that it is a bug. Information about this has been published here and the patach can be downloaded from here.

SP2 is out. Did you download new samples?

Thu, 22 Feb 2007 04:00:09 GMT

You must have already downloaded the SP2 now. If not, here is the link. Note that new samples are not included with it. Samples can be downloaded from here and readme file from here if want.

Did you notice that there is a new sample database called AdventureWorksLT?

Wanna use c# component in SSIS?

Sun, 18 Feb 2007 06:32:51 GMT

If you make search on it, you will definitely find enough information related to it. But since I experienced it, thought to blog. Assume that we have a .net class that is written with c#.net. Now I want to use one of the methods in that class in my SSIS package. Simply, I add “Script Task” and open the "Design Script" window by using edit menu item of it.

As usual, I have to reference the assembly, open the "Add Rererence" dialog box. My assembly is located in one my working folder; need to browse it for adding. But I found NO browse button and my assembly is not listed too. The reason is, I can use only assemblises located in "C:\WINDOWS\Microsoft.NET\Framework\v2.x.xxxxx" folder, means, I have to copy my assembly into that folder. Once copied, I referenced and design-time availability is okay.

Now the run-time. Can I run without any problem? No, I get "Could not load file or assembly....." error. Now the reason is, if SSIS accesses any assemblies, they should be exist in the GAC. Once my assembly is added to the GAC (of course, it need to be signed), it started working as I want; I can access the .net classes through SSIS.

SSIS: Some of foreach items are missing....

Wed, 31 Jan 2007 01:12:01 GMT

It was noticed when I opened one of my SSIS packages that had ForEachLoopContainer used. The exclamation mark was appeared on the ForEachLoopContainer and found out the reason for it was missing Foreach File Enumerator item. Not only that, the Foreach Item Enumerator was missing too. How did it happen? As usual, made a search and found out that it is a known issue, introduced by SP1. The problem has been discussed under below link and some contain possible solution.

http://forums.microsoft.com/MSDN/ShowPost.aspx?PostID=1146061&SiteID=1
support.microsoft.com/default.aspx?scid=kb;en-us;913817
http://mohansmindstorms.spaces.live.com/Blog/cns!69AE1BEA50F1D0E7!203.entry

Unfortunately, nothing worked for me. I changed the registry, re-installed, etc.. I would have corrected it if have had a backup of the registry because one possible reason for this is, registry changes made by other newly installed software. I suspect that culprit is WindowsDesktopSearchEngine, that I tried to install for my outlook 2007 but failed. So, the whole rescue operation ended up with re-installing the OS. How bad?

So, guys be aware of this issue.

SELECT * FROM: Does it always a bad practice?

Thu, 25 Jan 2007 07:18:40 GMT

We all know that “SELECT *” is a bad practice and it is expensive. But does this “theory” apply for all scenarios? No it does not. There is one place we should not worry about this. See the code below.

IF EXISTS (SELECT * FROM Person.Address)
BEGIN
print 'yes'
END

If “STATISTICS IO” sets on and executes the above code, you can see that only few logical or physical pages have been read. But if “SELECT * FROM Person.Address” statement executes without “EXIST”, all the pages for the table will be read. This shows that “SELECT *” is not an expensive query with “EXIST” because SQL Server check the availability of records instead reading the entire table. So, do not worry about “SELECT” statements with “EXIST”, SQL Server completely ignores the statement.

Using QUOTENAME for parameter values

Tue, 16 Jan 2007 05:17:59 GMT

I was going through few notes I had made regarding sql injection couple of years ago and it remembered me the value of some good practices when we write dynamic queries. Using QUOTENAME function for parameter values was one of among. See the below code for brushing up your mind too.

Assume that we wanted to have a stored procedure that returns some records from one of our business tables base on the column name (from a drop-down) and the value for it. You might be writing the SP like this;

CREATE PROCEDURE TestProc @ColumnName varchar(100), @Value varchar(100)
AS
BEGIN

   DECLARE @sql varchar(1000)
   SET @sql = 'SELECT columns FROM table WHERE '
   SET @sql = @sql + @ColumnName + ' = ''' + @Value + ''''

   EXEC (@sql)
END

So the client will execute this with;

DECLARE @ColumnName varchar(100), @Value varchar(100)
SET @ColumnName = 'Column1'
SET @Value = 'value'
EXEC TestProc @ColumnName, @Value

He got the result he wanted. What happen if he/she set the @value like this?

SET @Value = '''; SELECT * FROM sys.sysobject --'

Too bad. He got something that he should not see. Here is the worse case.

SET @Value = '''; DELETE FROM table1 --'

So, where we can use QUOTENAME function? Here is the way. Modify the SP like below.

SET @sql = @sql + @ColumnName + ' = ' + QUOTENAME(@Value, '''')

Keep this in your mind. Even though it is not a good practice writing dynamic SQLs, some situations force you to apply dynamic queries. If so, make sure you use this function too.

Service Pack 2 - December CTP

Thu, 21 Dec 2006 04:01:38 GMT

SQL Server 2005 Servise Pack 2 - Community Technolodgy Perview is now avaialble for download. Download from here.

Why cannot I enter date earlier than 1753?

Wed, 15 Nov 2006 16:46:44 GMT

Almost all SQL Server database developers know this issue, in other word, limitation. You cannot insert values lower than 01/01/1753 into datetime column. Do you know the reason? I was unaware of this reason but yesterday, one of my friends (He insisted not mention his name :)) explained me one of possible reasons. And guess what, it is not a technical reason, it is historical reason.

You might know that there had been two calendar types; Julian and Gregorian. When the people who used Julian decided move to Gregorian, they had to skip couple of days (10 to 14 days). This was happen around 1752. Because of this, the dates earlier than 1753 are different from country to country and have to maintain different procedures for different countries (or cultures) if the system need to support for them. That’s why SQL Server treats 01/01/1753 as the earliest date.

vardecimal with SP2: It is not what I thought

Sat, 11 Nov 2006 16:10:40 GMT

Is new datatype introduced? Is it just like varchar? I thought yes. But it is not. Vardecimal is a storage format, not a datatype. Because of that it is not available for columns when design tables. I have been trying to find what it is, yes after Gogula's comment for my one of previous posts. Finally I figured it out how to use it; see the sample code I used for testing.

-- create a table with decimal column
CREATE TABLE TestTable
(ID int IDENTITY(1,1) PRIMARY KEY, Number decimal(20,3))
GO

-- insert some values
DECLARE @i int
SET @i = 1
WHILE @i < 100000
BEGIN

   INSERT INTO TestTable
   VALUES (rand()*1000000)

   SET @i = @i + 1
END

-- check the space used for table
sp_spaceused 'TestTable'

-- result
-- reserved: 2632KB
-- data: 2576KB

Since vardecimal is a storage format, it cannot be set it to column itself. It has to be set for; first database and then table. But before all of these, it is possible to check whether the size reduction after enabling the format is really a worth enough.

-- check the estimated reduction for the row.
sys.sp_estimated_rowsize_reduction_for_vardecimal 'TestTable'

*Note: BOL shows the name of SP as sys.sp_estimate_ro... . Make sure you use it as sys.sp_estimated_ro.... .

-- result
-- avg_rowlen_fixed_format: 24.00
-- avg_rowlen_vardecimal_format: 17.98

The result tells us that the row size can be reduced by 25%. Isn't good? Yes, It is worthwhile to change the storage format of decimals. Enable it on the database and then the table.

-- enable the database for vardecimal storage format
sp_db_vardecimal_storage_format 'TestDatabase', 'ON'

-- enable the table
sp_tableoption 'TestTable', 'vardecimal storage format', true

Done. Lets check the space usage again.

-- check the space used for table
sp_spaceused 'TestTable'

-- result
-- reserved: 2248KB
-- data: 2176KB

Great. We just saved the disk space. And this is just for 100000 records. If you have millions of records with decimal values, Then it is high time to set this. Note that it is better to check the size reduction estimate before enabling it on the table because it might give you an unexpected result.