Q & A

What is a self join? Explain it with an example.
A.Self join is just like any other join, except that two instances of the same table will be joined in the query. Here is an example: Employees table which contains rows for normal employees as well as managers. So, to find out the managers of all the employees, you need a self join.

CREATE TABLE emp
(
empid int,
mgrid int,
empname char(10)
)

INSERT emp SELECT 1,2,'Vyas'
INSERT emp SELECT 2,3,'Mohan'
INSERT emp SELECT 3,NULL,'Shobha'
INSERT emp SELECT 4,2,'Shridhar'
INSERT emp SELECT 5,2,'Sourabh'

SELECT t1.empname [Employee], t2.empname [Manager]
FROM emp t1, emp t2
WHERE t1.mgrid = t2.empid

Here's an advanced query using a LEFT OUTER JOIN that even returns the employees without managers (super bosses)

SELECT t1.empname [Employee], COALESCE(t2.empname, 'No manager') [Manager]
FROM emp t1
LEFT OUTER JOIN
emp t2
ON
t1.mgrid = t2.empid


There is a trigger defined for INSERT operations on a table, in an OLTP system. The trigger is written to instantiate a COM object and pass the newly insterted rows to it for some custom processing. What do you think of this implementation? Can this be implemented better?
A.Instantiating COM objects is a time consuming process and since you are doing it from within a trigger, it slows down the data insertion process. Same is the case with sending emails from triggers. This scenario can be better implemented by logging all the necessary data into a separate table, and have a job which periodically checks this table and does the needful.

What are triggers? How many triggers you can have on a table? How to invoke a trigger on demand?
A.Triggers are special kind of stored procedures that get executed automatically when an INSERT, UPDATE or DELETE operation takes place on a table.
In SQL Server 6.5 you could define only 3 triggers per table, one for INSERT, one for UPDATE and one for DELETE. From SQL Server 7.0 onwards, this restriction is gone, and you could create multiple triggers per each action. But in 7.0 there's no way to control the order in which the triggers fire. In SQL Server 2000 you could specify which trigger fires first or fires last using sp_settriggerorder
Triggers can't be invoked on demand. They get triggered only when an associated action (INSERT, UPDATE, DELETE) happens on the table on which they are defined.
Triggers are generally used to implement business rules, auditing. Triggers can also be used to extend the referential integrity checks, but wherever possible, use constraints for this purpose, instead of triggers, as constraints are much faster.
Till SQL Server 7.0, triggers fire only after the data modification operation happens. So in a way, they are called post triggers. But in SQL Server 2000 you could create pre triggers also. Search SQL Server 2000 books online for INSTEAD OF triggers.
Also check out books online for 'inserted table', 'deleted table' and COLUMNS_UPDATED()

What is the system function to get the current user's user id?
A.USER_ID(). Also check out other system functions like USER_NAME(), SYSTEM_USER, SESSION_USER, CURRENT_USER, USER, SUSER_SID(), HOST_NAME().

What is an extended stored procedure? Can you instantiate a COM object by using T-SQL?
A.An extended stored procedure is a function within a DLL (written in a programming language like C, C++ using Open Data Services (ODS) API) that can be called from T-SQL, just the way we call normal stored procedures using the EXEC statement.
Yes, you can instantiate a COM (written in languages like VB, VC++) object from T-SQL by using sp_OACreate stored procedure. Also see books online for sp_OAMethod, sp_OAGetProperty, sp_OASetProperty, sp_OADestroy. For an example of creating a COM object in VB and calling it from T-SQL, see 'My code library' section of this site.

Can you have a nested transaction?
A.Yes, very much. Check out BEGIN TRAN, COMMIT, ROLLBACK, SAVE TRAN and @@TRANCOUNT

What is a join and explain different types of joins.
A.Joins are used in queries to explain how different tables are related. Joins also let you select data from a table depending upon data from another table.

Types of joins: INNER JOINs, OUTER JOINs, CROSS JOINs. OUTER JOINs are further classified as LEFT OUTER JOINS, RIGHT OUTER JOINS and FULL OUTER JOINS.

Write down the general syntax for a SELECT statements covering all the options.
A.Here's the basic syntax:
SELECT select_list
[INTO new_table_]
FROM table_source
[WHERE search_condition]
[GROUP BY group_by_expression]
[HAVING search_condition]
[ORDER BY order_expression [ASC | DESC] ]

What are cursors? Explain different types of cursors. What are the disadvantages of cursors? How can you avoid cursors?
A.Cursors allow row-by-row prcessing of the resultsets.

Types of cursors: Static, Dynamic, Forward-only, Keyset-driven. See books online for more information.
Disadvantages of cursors: Each time you fetch a row from the cursor, it results in a network roundtrip, where as a normal SELECT query makes only one rowundtrip, however large the resultset is. Cursors are also costly because they require more resources and temporary storage (results in more IO operations). Furthere, there are restrictions on the SELECT statements that can be used with some types of cursors.
Most of the times, set based operations can be used instead of cursors. Here is an example:
If you have to give a flat hike to your employees using the following criteria:
Salary between 30000 and 40000 -- 5000 hike
Salary between 40000 and 55000 -- 7000 hike
Salary between 55000 and 65000 -- 9000 hike
In this situation many developers tend to use a cursor, determine each employee's salary and update his salary according to the above formula. But the same can be achieved by multiple update statements or can be combined in a single UPDATE statement as shown below:
UPDATE tbl_emp SET salary =
CASE WHEN salary BETWEEN 30000 AND 40000 THEN salary + 5000
WHEN salary BETWEEN 40000 AND 55000 THEN salary + 7000
WHEN salary BETWEEN 55000 AND 65000 THEN salary + 10000
END
Another situation in which developers tend to use cursors: You need to call a stored procedure when a column in a particular row meets certain condition. You don't have to use cursors for this. This can be achieved using WHILE loop, as long as there is a unique key to identify each row.

What is database replicaion? What are the different types of replication you can set up in SQL Server?
A.Replication is the process of copying/moving data between databases on the same or different servers. SQL Server supports the following types of replication scenarios:

• Snapshot replication
• Transactional replication (with immediate updating subscribers, with queued updating subscribers)
• Merge replication
  

Explian different types of BACKUPs avaialabe in SQL Server? Given a particular scenario, how would you go about choosing a backup plan?
A.Types of backups you can create in SQL Sever 7.0+ are Full database backup, differential database backup, transaction log backup, filegroup backup.

What are the different ways of moving data/databases between servers and databases in SQL Server?
A.There are lots of options available, you have to choose your option depending upon your requirements. Some of the options you have are: BACKUP/RESTORE, dettaching and attaching databases, replication, DTS, BCP, logshipping, INSERT...SELECT, SELECT...INTO, creating INSERT scripts to generate data.

What are statistics, under what circumstances they go out of date, how do you update them
A.Statistics determine the selectivity of the indexes. If an indexed column has unique values then the selectivity of that index is more, as opposed to an index with non-unique values. Query optimizer uses these indexes in determining whether to choose an index or not while executing a query.

Some situations under which you should update statistics:
1) If there is significant change in the key values in the index
2) If a large amount of data in an indexed column has been added, changed, or removed (that is, if the distribution of key values has changed), or the table has been truncated using the TRUNCATE TABLE statement and then repopulated
3) Database is upgraded from a previous version

Look up SQL Server books online for the following commands: UPDATE STATISTICS, STATS_DATE, DBCC SHOW_STATISTICS, CREATE STATISTICS, DROP STATISTICS, sp_autostats, sp_createstats, sp_updatestats

As a part of your job, what are the DBCC commands that you commonly use for database maintenance?
A.DBCC CHECKDB, DBCC CHECKTABLE, DBCC CHECKCATALOG, DBCC CHECKALLOC, DBCC SHOWCONTIG, DBCC SHRINKDATABASE, DBCC SHRINKFILE etc. But there are a whole load of DBCC commands which are very useful for DBAs

How to restart SQL Server in single user mode? How to start SQL Server in minimal configuration mode?
A.SQL Server can be started from command line, using the SQLSERVR.EXE. This EXE has some very important parameters with which a DBA should be familiar with. -m is used for starting SQL Server in single user mode and -f is used to start the SQL Server in minimal confuguration mode. Check out SQL Server books online for more parameters and their explanations.

Explain CREATE DATABASE syntax
A.Many of us are used to craeting databases from the Enterprise Manager or by just issuing the command: CREATE DATABAE MyDB. But what if you have to create a database with two filegroups, one on drive C and the other on drive D with log on drive E with an initial size of 600 MB and with a growth factor of 15%? That's why being a DBA you should be familiar with the CREATE DATABASE syntax. Check out SQL Server books online for more information.

What is blocking and how would you troubleshoot it?
A.Blocking happens when one connection from an application holds a lock and a second connection requires a conflicting lock type. This forces the second connection to wait, blocked on the first.

What is a deadlock and what is a live lock? How will you go about resolving deadlocks?
A.Deadlock is a situation when two processes, each having a lock on one piece of data, attempt to acquire a lock on the other's piece. Each process would wait indefinitely for the other to release the lock, unless one of the user processes is terminated. SQL Server detects deadlocks and terminates one user's process.

A livelock is one, where a request for an exclusive lock is repeatedly denied because a series of overlapping shared locks keeps interfering. SQL Server detects the situation after four denials and refuses further shared locks. A livelock also occurs when read transactions monopolize a table or page, forcing a write transaction to wait indefinitely.


What are the steps you will take, if you are tasked with securing an SQL Server?
A.Again this is another open ended question. Here are some things you could talk about: Preferring NT authentication, using server, databse and application roles to control access to the data, securing the physical database files using NTFS permissions, using an unguessable SA password, restricting physical access to the SQL Server, renaming the Administrator account on the SQL Server computer, disabling the Guest account, enabling auditing, using multiprotocol encryption, setting up SSL, setting up firewalls, isolating SQL Server from the web server etc.
http://vyaskn.tripod.com/sql_server_security_best_practices.htm


What are the steps you will take to improve performance of a poor performing query?
A.This is a very open ended question and there could be a lot of reasons behind the poor performance of a query. But some general issues that you could talk about would be: No indexes, table scans, missing or out of date statistics, blocking, excess recompilations of stored procedures, procedures and triggers without SET NOCOUNT ON, poorly written query with unnecessarily complicated joins, too much normalization, excess usage of cursors and temporary tables.
Some of the tools/ways that help you troubleshooting performance problems are: SET SHOWPLAN_ALL ON, SET SHOWPLAN_TEXT ON, SET STATISTICS IO ON, SQL Server Profiler, Windows NT /2000 Performance monitor, Graphical execution plan in Query Analyzer.

What is RAID and what are different types of RAID configurations?
A.RAID stands for Redundant Array of Inexpensive Disks, used to provide fault tolerance to database servers. There are six RAID levels 0 through 5 offering different levels of performance, fault tolerance.
http://www.raid-advisory.com/

What is an index? What are the types of indexes? How many clustered indexes can be created on a table? I create a separate index on each column of a table. what are the advantages and disadvantages of this approach?
A.Indexes in SQL Server are similar to the indexes in books. They help SQL Server retrieve the data quicker.
Indexes are of two types. Clustered indexes and non-clustered indexes. When you craete a clustered index on a table, all the rows in the table are stored in the order of the clustered index key. So, there can be only one clustered index per table. Non-clustered indexes have their own storage separate from the table data storage. Non-clustered indexes are stored as B-tree structures (so do clustered indexes), with the leaf level nodes having the index key and it's row locater. The row located could be the RID or the Clustered index key, depending up on the absence or presence of clustered index on the table.
If you create an index on each column of a table, it improves the query performance, as the query optimizer can choose from all the existing indexes to come up with an efficient execution plan. At the same t ime, data modification operations (such as INSERT, UPDATE, DELETE) will become slow, as every time data changes in the table, all the indexes need to be updated. Another disadvantage is that, indexes need disk space, the more indexes you have, more disk space is used.

What's the difference between DELETE TABLE and TRUNCATE TABLE commands?
A.DELETE TABLE is a logged operation, so the deletion of each row gets logged in the transaction log, which makes it slow. TRUNCATE TABLE also deletes all the rows in a table, but it won't log the deletion of each row, instead it logs the deallocation of the data pages of the table, which makes it faster. Of course, TRUNCATE TABLE can be rolled back.

What is lock escalation?
A.Lock escalation is the process of converting a lot of low level locks (like row locks, page locks) into higher level locks (like table locks). Every lock is a memory structure too many locks would mean, more memory being occupied by locks. To prevent this from happening, SQL Server escalates the many fine-grain locks to fewer coarse-grain locks. Lock escalation threshold was definable in SQL Server 6.5, but from SQL Server 7.0 onwards it's dynamically managed by SQL Server.

CREATE INDEX myIndex ON myTable(myColumn)
What type of Index will get created after executing the above statement?
A.Non-clustered index. Important thing to note: By default a clustered index gets created on the primary key, unless specified otherwise.

Explain different isolation levels
A.An isolation level determines the degree of isolation of data between concurrent transactions. The default SQL Server isolation level is Read Committed. Here are the other isolation levels (in the ascending order of isolation): Read Uncommitted, Read Committed, Repeatable Read, Serializable. See SQL Server books online for an explanation of the isolation levels. Be sure to read about SET TRANSACTION ISOLATION LEVEL, which lets you customize the isolation level at the connection level.

What is a transaction and what are ACID properties?
A.A transaction is a logical unit of work in which, all the steps must be performed or none. ACID stands for Atomicity, Consistency, Isolation, Durability. These are the properties of a transaction.

What are defaults? Is there a column to which a default can't be bound?
A.A default is a value that will be used by a column, if no value is supplied to that column while inserting data. IDENTITY columns and timestamp columns can't have defaults bound to them.

Define candidate key, alternate key, composite key.
A candidate key is one that can identify each row of a table uniquely. Generally a candidate key becomes the primary key of the table. If the table has more than one candidate key, one of them will become the primary key, and the rest are called alternate keys.

A key formed by combining at least two or more columns is called composite key.

What is bit datatype and what's the information that can be stored inside a bit column?
A.Bit datatype is used to store boolean information like 1 or 0 (true or false). Untill SQL Server 6.5 bit datatype could hold either a 1 or 0 and there was no support for NULL. But from SQL Server 7.0 onwards, bit datatype can represent a third state, which is NULL.

What are user defined datatypes and when you should go for them?
A.User defined datatypes let you extend the base SQL Server datatypes by providing a descriptive name, and format to the database. Take for example, in your database, there is a column called Flight_Num which appears in many tables. In all these tables it should be varchar(8). In this case you could create a user defined datatype called Flight_num_type of varchar(8) and use it across all your tables.

What's the difference between a primary key and a unique key?
A.Both primary key and unique enforce uniqueness of the column on which they are defined. But by default primary key creates a clustered index on the column, where are unique creates a nonclustered index by default. Another major difference is that, primary key doesn't allow NULLs, but unique key allows one NULL only.

How you can rename a database ?
A. To rename a database first get the database into Single user mode . by using the query below ( run these query in query analyzer)

ALTER DATABASE DBMydb SET SINGLE_USER WITH ROLLBACK IMMEDIATE

then rename the database using sp_rename

sp_renamedb 'DBMydb', 'DBmydb_new'

then change the access mode of the database to multiuser

ALTER DATABASE DBmydb_new SET MULTI_USER

What is mean by suspect status of a database ?
A. SQL Server gives a database 'suspect' status when it is unable to recover the database . reasons may be many , may be a torn page detection , may be datafile location no found.

How many access mode of database are in MSSQL
A there are three access mode of a database SINGLE_USER , RESTRICTED_USER and MULTI_USER . SINGLE_USER mode restricts database access to one user Connection. RESTRICTED_USER mode restricts database access to db_owner, dbcreator, and sysadmin roles. All other users are disconnected.MULTI_USER mode lets all users with proper access connect to the database.

What are the default databases in MSSQL Server Instance after Fresh installation
A. there are 4 default databases .

• Master
  This is your most critical database, and is the core of your SQL Server implementation.
  Limit changes and access to this database as much as possible. Without the master
  database, your SQL Server instance will not start up.
• MSDB
  This database contains tables that control and track the elements of the SQL Server
  Agent, as well as replication, log shipping (Enterprise Edition), database backup and
  restore history, database maintenance plan configurations and history, and a subset of
  tables participating in replication status and jobs. DTS packages saved to SQL Server are also stored within this database.
• Model
  The model database is a template database, which, when modified, will affect the default
  values of databases that are created in the future. For example, if a table is added to the
  model database, future databases created for the SQL Server instance will include this
  table.
• Tempdb
  The tempdb database is used to hold all temporary tables and temporary stored
  procedures. SQL Server also uses it for internal sorting and query operations (called
  work tables, or spooling). This database is re-created each time SQL Server is started,
  and cannot be used to store permanent tables.

What is replication ?
A. Replication is a process which enables data and database objects to be copied and modified from one database to another across different networks and platforms.the physical seperation of the databases and latency are integral part of the design process in replication. MSSQL Server 2000 permits 3 different kind of replication. they are snapshot,transactional and merge.

How many typf of backups are supported by MSSQL ?
A. there are three type of backup

• Full Database Backup.
• Transaction Log Backup
• Differential Backup
  

Whats is Log Shipping ?
A. Log shipping is the process of backing up of database and transaction log files on a production SQL server, and then restoring them onto a standby server. The key feature of log shipping is that it will automatically backup transaction logs throughout the day (for whatever interval you specify) and automatically restore them on the standby server. This in effect keeps the two SQL Servers in "sync". Should the production server fail, all you have to do is point the users to the new server, and you are all set.

Name the types of Recovery model in SQL Server ?
A. There are thre recovery model in MSSQL Server
• Simple Recovery Model .
• Full Recovery Model.
• Bulk Logged Recovery Model.

What is a Dead Lock ?
A . Deadlock occurs when two users have locks on separate objects and each user wants a lock on the other's object. For example, User1 has a lock on object "A" and wants a lock on object "B" and User2 has a lock on object "B" and wants a lock on object "A". In this case, SQL Server ends a deadlock by choosing the user, who will be a deadlock victim. After that, SQL Server rolls back the breaking user's transaction, sends message number 1205 to notify the user's application about breaking, and then allows the nonbreaking user's process to continue.
Note. To reduce the chance of a deadlock, you should minimize the size of transactions and transaction times.

How many type of Locks are there ?
A. There are three main types of locks that SQL Server 7.0/2000 uses
Shared locks
Update locks
Exclusive locks

A database is consists of how many type of files ? Name their extensions
A. A database can consist of three types of file :
Primary file , Seconday file and transaction logs
The primary data file is the starting point of the database and points to any other files in the database. If the primary file is not large enough to hold all the data in the database, then secondary files are required.Transaction logs files hold records of updates to the data files, and can be used to recover the database after an error. Every database has at least one transaction log file, and some databases may have more than one.
The recommended file name extension for primary data files is .mdf.
The recommended file name extension for secondary data files is .ndf.
The recommended file name extension for log files is .ldf

How many type of indexes are there ? How many Index can be created on a table ?
A. there are two type of indexes Clustered and Non-Clustured, 249 Non Clustured and 1 Clustered index

How you can get the last identity value inserted in any table ?
A. SQL Server has a System Variable @@IDENTITY which gives the last identity element value inserted in any table

How many type of Contraints are in MSSQL.
A. SQL Server 2000 supports five classes of constraints.
1) NOT NULL
2) CHECK
3) UNIQUE
4) PRIMARY KEY
5) FOREIGN KEY

How you can trap error in Sql Server
A. by using @@error .
After each Transact-SQL statement, the server sets the variable to an integer value:
0—if the statement was successful
Error number—if the statement has failed

What is @@Fetch_status ?
A. @@fetch_status is a function (or global variable) that returns the success code of the last Fetch statement executed during the current connection. It is often used as an exit criterion in loops that fetch records from a cursor.

Whats the limitation of user defined funtion ?
A User-defined functions have one serious limitation. They cannot have side effects. A function side effect is any permanent change to resources (such as tables) that have a scope outside of the function (such as a non-temporary table that is not declared in the function). Basically, this requirement means that a function should return a value while changing nothing in the database. it means in short that "user defined function can not use UPDATE / DELETE on permament table objects in MSSQL"

In how many ways you can recieve information from stored procedure
A. there are 4 ways to receive information from a stored procedure:
Resultset , Parameters , Return value , A global cursor that can be referenced outside the stored procedure.

How many type of triggers are there in Sql Server.
A. There are two type of triggers
• After Triggers
• Instead of Triggers

Whats the Difference between datetime and smalldatetime datatype in SQL Server .
A . The main difference between these two datatypes is in the amount of space they occupy. datetime occupies eight bytes and smalldatetime only four. The difference in size is due to a difference in precision. The precision of smalldatetime is one minute, and it covers dates from January 1, 1900 , through June 6, 2079 , which is usually more than enough. The precision of datetime is 3.33 ms, and it covers dates from January 1, 1753 , to December 31, 9999 .

How many type of authentication method are there in SQL Server 2005 and 2000 and what is that ?
A. There are two type of authentication method in SQL Server 2005 and 2000
Windows Authentication
SQL Server Authentication

How can you generate GUID in in Transact-SQL ?
A. GUIDs can be generated using the NEWID function.

How many system datatypes are in SQL Server.
A. 27

What is the Maximum number of input and output parameters in Stored procedure in SQL Server 2000 ?
A. 1024

How can we improve stored procedure performance? Explain the performance tuning techniques.

1. Use many WHERE clauses in the SELECT statements.
2. Select only those fields which you really require.
3. Joins are expensive in terms of time. Join the tables using related fields manly indexed fields.
4. Don't use unused tables.

NOTES: When you execute the stored proc, SQL Server generates an execution plan based on your indexes and statistics available. Now it is possible that the stored proc inserts a lot of rows into some table which is used in another statement in a join and the plan that was built is not efficient for the subsequent sql statement because the profile of the table has changed since the plan was generated. I would suggest splitting the stored proc into two and the split should be done such that the second proc deals with the updated table. When the second proc is executed, it will have a more realistic execution plan for the updated table.

TRY...CATCH in SQL Server 2005

Introduction
SQL Server 2005 offers a number of new features over its predecessor, including many features aimed at making working with databases more like writing .NET application code. For example, in SQL Server 2005, stored procedures, triggers, UDFs, and so on can be written using any .NET Framework programming language (such as Visual Basic or C#). Another feature, and the focus of this article, is SQL Server 2005's support for TRY...CATCH blocks. TRY...CATCH blocks are the standard approach to exception handling in modern programming languages, and involve:

• A TRY Block - the TRY block contains the instructions that might cause an exception
• A CATCH Block - if an exception occurs from one of the statements in the TRY block, control is branched to the CATCH block, where the exception can be handled, logged, and so on.

See the Wikipedia Exception Handling entry for more information on the TRY...CATCH construct as well as exception handling concepts in general.

Prior to SQL Server 2005, detecting errors resulting from T-SQL statements could only be handled by checking a global error variable, @@ERROR. Because the @@ERROR variable value is reset after each SQL statement, this antiquated approach leads to rather bloated stored procedures, as the variable must be checked after each statement with code to handle any problems.

The TRY...CATCH block in SQL Server 2005 offers a much more readable syntax and one that developers are more familiar with. In this article we'll look at the new TRY...CATCH block and examine how it can be used to rollback a transaction in the face of an error. Read on to learn more!

Checking @@ERROR - the Old Way of Handling Errors in T-SQL
SQL Server provides an @@ERROR variable that indicates the status of the last completed SQL statement in a given batch. If a SQL statement is completed successfully, @@ERROR is assigned 0. If, however, an error occurs, @@ERROR is set to the number of the error message.

To see how the @@ERROR variable can be used, imagine that we have a data-driven web application that maintains employee information. Let's assume that our database has Employees and EmployeePhoneNumbers tables, among others. These two tables share a one-to-many relationship; that is, each Employees record can have an arbitrary number of related records in the EmployeePhoneNumbers table. There might be one for their office phone, one for their pager, one for their cell phone, and so on. Imagine that our database includes a stored procedure, DeleteEmployee, which is comprised of two DELETE statements - one to delete the employee's related phone numbers from the system and one to delete the actual employee record:

CREATE PROCEDURE DeleteEmployee ( @EmployeeID int ) AS -- Delete the Employee's phone numbers DELETE FROM EmployeePhoneNumbers WHERE EmployeeID = @EmployeeID -- Delete the Employee record DELETE FROM Employees WHERE EmployeeID = @EmployeeID

Since we want these two delete statements to be atomic and either both fail or both succeed, we need to wrap up these two statements into a transaction. By using a transaction, we can rollback the transaction in the face of an error and undo any changes made since the start of the exception. To accomplish this we might initially try to use the following syntax:

CREATE PROCEDURE DeleteEmployee ( @EmployeeID int ) AS BEGIN TRANSACTION -- Start the transaction -- Delete the Employee's phone numbers DELETE FROM EmployeePhoneNumbers WHERE EmployeeID = @EmployeeID -- Delete the Employee record DELETE FROM Employees WHERE EmployeeID = @EmployeeID -- See if there is an error IF @@ERROR <> 0 -- There's an error b/c @ERROR is not 0, rollback ROLLBACK ELSE COMMIT -- Success! Commit the transaction

This stored procedure (it appears) starts a transaction, runs the two DELETE statements, and then checks to see if there was an error. If there was one, it rolls the transaction back, else it commits the transaction. I say "it appears" because this syntax, while legal, is semantically incorrect because the @@ERROR variable is set after every SQL statement. Therefore, if the first DELETE statement has an error the @@ERROR variable will be set to its error number. Then, the second DELETE will execute. If this second DELETE succeeds, @@ERROR will be set back to 0, in which case the transaction will be committed even though there was a problem with the first statement! Whoops!

Instead, a check must be made after every SQL statement to see if there has been an error. If so, the transaction must be rolled back and the stored procedure exited. This can lead to bulky scripts as a stored procedure with, say, five statements will have five checks against the @@ERROR variable. And if you forget to cut and paste a check in for a particular statement you're opening yourself up to a potential problem.

For more information on transactions and the @@ERROR syntax used for checking for errors and rolling back as needed, see Managing Transactions in SQL Server Stored Procedures.

Handling Errors With SQL Server 2005's TRY...CATCH Blocks
While SQL Server 2005 still supports the @@ERROR approach, a better alternative exists with its new TRY...CATCH blocks. As with programming languages like Visual Basic, C#, and Java, the SQL Server 2005 TRY...CATCH block executes a number of statements in the TRY block. If there are no errors in any of the statements, control proceeds to after the CATCH block. If, however, one of the statements causes an error, control branches immediately to the start of the CATCH block. Furthermore, like programming languages, nested TRY...CATCH blocks are allowed, meaning that you can have an entire TRY...CATCH block in the TRY or CATCH portions of an "outter" TRY...CATCH block.

BEGIN TRY Try Statement 1 Try Statement 2 ... Try Statement M END TRY BEGIN CATCH Catch Statement 1 Catch Statement 2 ... Catch Statement N END CATCH

The following system functions are available in the CATCH block and can be used to determine additional error information:

Function	Description
ERROR_NUMBER()	Returns the number of the error
ERROR_SEVERITY()	Returns the severity
ERROR_STATE()	Returns the error state number
ERROR_PROCEDURE()	Returns the name of the stored procedure or trigger where the error occurred
ERROR_LINE()	Returns the line number inside the routine that caused the error
ERROR_MESSAGE()	Returns the complete text of the error message. The text includes the values supplied for any substitutable parameters, such as lengths, object names, or times

Also realize that not all errors generating by the TRY block statements are passed onto the CATCH block. Any errors with a severity of 10 or less are considered to be warnings and do not branch control flow to the CATCH block. Also, any errors that sever the database connection will not cause the CATCH block to be reached.

Let's look at a quick example of using TRY...CATCH, after which we'll turn our attention to using this new construct for rolling back transactions in the face of an error. The following example shows a very simply INSERT query on the Northwind database's Products table. The Products table's ProductID column is an IDENTITY column and therefore its value can't be specified when inserting a new record. However, I've specified this value in the following INSERT statement. Hence, control is turned over to the CATCH block where error information is displayed.

BEGIN TRY -- This will generate an error, as ProductID is an IDENTITY column -- Ergo, we can't specify a value for this column... INSERT INTO Products(ProductID, ProductName) VALUES(1, 'Test') END TRY BEGIN CATCH SELECT 'There was an error! ' + ERROR_MESSAGE() END CATCH

This query will return a single record with a single column with the contents: "There was an error! Cannot insert explicit value for identity column in table 'Products' when IDENTITY_INSERT is set to OFF."

Using TRY...CATCH to Rollback a Transaction in the Face of an Error
As discussed earlier in this article, one of the downsides of the @@ERROR variable approach is that for transactions a check against this variable must be added after each and every SQL statement to determine if an error occurred and, if so, to rollback the transaction. With SQL Server 2005's TRY...CATCH block, however, these types of scripts are greatly simplified, as the following example illustrates:

CREATE PROCEDURE DeleteEmployee ( @EmployeeID int ) AS BEGIN TRY BEGIN TRANSACTION -- Start the transaction -- Delete the Employee's phone numbers DELETE FROM EmployeePhoneNumbers WHERE EmployeeID = @EmployeeID -- Delete the Employee record DELETE FROM Employees WHERE EmployeeID = @EmployeeID -- If we reach here, success! COMMIT END TRY BEGIN CATCH -- Whoops, there was an error IF @@TRANCOUNT > 0 ROLLBACK -- Raise an error with the details of the exception DECLARE @ErrMsg nvarchar(4000), @ErrSeverity int SELECT @ErrMsg = ERROR_MESSAGE(), @ErrSeverity = ERROR_SEVERITY() RAISERROR(@ErrMsg, @ErrSeverity, 1) END CATCH

In the TRY block a transaction is started and the two DELETE statements are performed. If both DELETEs succeed, the COMMIT will be reached and the transaction committed. If, however, either one produces an error, control will be routed to the CATCH block where the transaction will be rolled back. Also, the CATCH block re-raises the error (using RAISERROR) so that the error information will be percolated up to the application that invoked the stored procedure. For an ASP.NET web application, that means that an exception will be raised in the .NET code that invoked this stored procedure, as chances are you not only want to rollback the transaction, but have some sort of error message handled in the web application as well so that the end user knows that their action failed.

Adding the call to RAISERROR in the CATCH block is tantamount to having a TRY...CATCH block in a programming language that re-throws the exception in the CATCH block after logging it or performing other actions. The action performed in the example above is rolling back the transaction, but could also include logging logic. If you omit the RAISERROR, the ASP.NET application won't throw an exception from executing the database command. Chances are you want to have an exception thrown on the ASP.NET side (so that you don't fail silently). If so, leave in the RAISERROR call.

Conclusion
SQL Server 2005's new TRY...CATCH block brings the familiar TRY...CATCH exception handling to T-SQL. Prior to SQL Server 2005, errors could only be detected in SQL scripts through the use of the @@ERROR variable, which annoyingly reset after each SQL statement, thereby requiring checks after each and every statement. This led to bloated script that was prone to typos or cut and paste errors leading to potentially serious problems. With SQL Server 2005's TRY...CATCH block, on the other hand, anytime an error is raised by one of the statements in the TRY block, control is directed to the CATCH block. As we saw in this article, the TRY...CATCH block allows for much more readable and cleaner error handling in transaction settings.

Common Table Expressions (CTE) in SQL Server 2005

When crafting a query in SQL, there are often times when we may need to operate over a set of data that doesn't inherently exist within the system. For example, the database for an eCommerce web application would have the standard tables - Products, Customers, Orders, OrderDetails, and so on - but we may need to run reports on a particular subset of the data or against aggregate data across these tables. Or the reporting queries we need might need to group or filter by results returned by scalar subqueries. Typically, views are used to break down complex queries into digestible chunks or to provide scalar subquery results that can be grouped and filtered. Views, however, are sometimes overkill, as they are permanent objects at the system-level. If we only need to reference this complex query in a single stored procedure or UDF, another option is to use a derived table. Unfortunately, derived tables muddle the readability of the query and must be repeated for each use in a statement.

Common Table Expressions, or CTE, are a new construct introduced in Microsoft SQL Server 2005 that offer a more readable form of the derived table that can be declared once and referenced multiple times in a query. Moreover, CTEs can be recursively defined, allowing a recursive entity to be enumerated without the need for recursive stored procedures. In this article we'll examine the benefits, uses, and syntax of both recursive and non-recursive CTEs. Read on to learn more!

A Simple Common Table Expression Example
Before we dive into the syntax or gritty details of CTEs, let's start by looking at a simple example. I think you'll agree that even without knowing the syntax of CTEs, they are pretty readable and straightforward (the hallmark of a well-designed programming language construct).

WITH ProductAndCategoryNamesOverTenDollars (ProductName, CategoryName, UnitPrice) AS ( SELECT p.ProductName, c.CategoryName, p.UnitPrice FROM Products p INNER JOIN Categories c ON c.CategoryID = p.CategoryID WHERE p.UnitPrice > 10.0 ) SELECT * FROM ProductAndCategoryNamesOverTenDollars ORDER BY CategoryName ASC, UnitPrice ASC, ProductName ASC

This query creates a CTE named ProductAndCategoryNamesOverTenDollars that returns the name, category name, and price of those products whose unit price exceeds $10.00. Once the CTE has been defined, it must then immediately be used in a query. The query treats the CTE as if were a view or table in the system, returning the three fields defined by the CTE (ProductName, CategoryName, and UnitPrice), ordered alphabetically by category, then by price, and then alphabetically by product name.

In short, a Common Table Expression allows us to define a temporary, view-like construct. We start by (optionally) specifying the columns it returns, then define the query. Following that, the CTE can be used in a SELECT, INSERT, UPDATE, or DELETE statement.

Common Table Expression Syntax
A Common Table Expression contains three core parts:

• The CTE name (this is what follows the WITH keyword)
• The column list (optional)
• The query (appears within parentheses after the AS keyword)

The query using the CTE must be the first query appearing after the CTE. That is, you cannot do the following:

WITH ProductAndCategoryNamesOverTenDollars (ProductName, CategoryName, UnitPrice) AS ( SELECT p.ProductName, c.CategoryName, p.UnitPrice FROM Products p INNER JOIN Categories c ON c.CategoryID = p.CategoryID WHERE p.UnitPrice > 10.0 ) SELECT * FROM Products SELECT * FROM ProductAndCategoryNamesOverTenDollars ORDER BY CategoryName ASC, UnitPrice ASC, ProductName ASC

The ProductAndCategoryNamesOverTenDollars CTE only applies to the first query following it. So when the second query is reached, ProductAndCategoryNamesOverTenDollars is undefined, resulting in an "Invalid object name 'ProductAndCategoryNamesOverTenDollars'" error message.

You can, however, define multiple CTEs after the WITH keyword by separating each CTE with a comma. For example, the following query uses two CTEs. The subsequent SELECT query then uses an INNER JOIN to match together the records from the two CTEs:

WITH CategoryAndNumberOfProducts (CategoryID, CategoryName, NumberOfProducts) AS ( SELECT CategoryID, CategoryName, (SELECT COUNT(1) FROM Products p WHERE p.CategoryID = c.CategoryID) as NumberOfProducts FROM Categories c ), ProductsOverTenDollars (ProductID, CategoryID, ProductName, UnitPrice) AS ( SELECT ProductID, CategoryID, ProductName, UnitPrice FROM Products p WHERE UnitPrice > 10.0 ) SELECT c.CategoryName, c.NumberOfProducts, p.ProductName, p.UnitPrice FROM ProductsOverTenDollars p INNER JOIN CategoryAndNumberOfProducts c ON p.CategoryID = c.CategoryID ORDER BY ProductName

Unlike a derived table, CTEs can be defined just once, yet appear multiple times in the subsequent query. To demonstrate this, consider the following example: the Northwind database's Employees table contains an optional ReportsTo column that, if specified, indicates the employee's manager. ReportsTo is a self-referencing foreign key, meaning that, if provided, it refers back to another EmployeeID in the Employees table. Imagine that we wanted to display a list of employees including how many other employees they directly managed. This could be done using a simple, CTE-free SELECT statement, but let's use a CTE for now (for reasons which will become clear soon):

WITH EmployeeSubordinatesReport (EmployeeID, LastName, FirstName, NumberOfSubordinates, ReportsTo) AS ( SELECT EmployeeID, LastName, FirstName, (SELECT COUNT(1) FROM Employees e2 WHERE e2.ReportsTo = e.EmployeeID) as NumberOfSubordinates, ReportsTo FROM Employees e ) SELECT LastName, FirstName, NumberOfSubordinates FROM EmployeeSubordinatesReport

This query will return the employees records, showing each employee's last and first name along with how many other employees they manage. As the figure below shows, only Andrew Fuller and Steven Buchanan are manager material.

Now, imagine that our boss (Andrew Fuller, perhaps) comes charging into our office and demands that the report also lists each employee's manager's name and number of subordinates (if the employee has a manager, that is - Mr. Fuller is all to quick to point out that he reports to no one). Adding such functionality is a snap with the CTE - just add it in a LEFT JOIN!

WITH EmployeeSubordinatesReport (EmployeeID, LastName, FirstName, NumberOfSubordinates, ReportsTo) AS ( SELECT EmployeeID, LastName, FirstName, (SELECT COUNT(1) FROM Employees e2 WHERE e2.ReportsTo = e.EmployeeID) as NumberOfSubordinates, ReportsTo FROM Employees e ) SELECT Employee.LastName, Employee.FirstName, Employee.NumberOfSubordinates, Manager.LastName as ManagerLastName, Manager.FirstName as ManagerFirstName, Manager.NumberOfSubordinates as ManagerNumberOfSubordinates FROM EmployeeSubordinatesReport Employee LEFT JOIN EmployeeSubordinatesReport Manager ON Employee.ReportsTo = Manager.EmployeeID

With this additional LEFT JOIN, the employee's manager's results are returned; if there's no manager for the employee, NULLs are returned instead.

When to Use Common Table Expressions

Common Table Expressions offer the same functionality as a view, but are ideal for one-off usages where you don't necessarily need a view defined for the system. Even when a CTE is not necessarily needed (as when listing just the employees and their subordinate count in the example above), it can improve readability. In Using Common Table Expressions, Microsoft offers the following four advantages of CTEs: Create a recursive query. Substitute for a view when the general use of a view is not required; that is, you do not have to store the definition in metadata. Enable grouping by a column that is derived from a scalar subselect, or a function that is either not deterministic or has external access. Reference the resulting table multiple times in the same statement. Using a CTE offers the advantages of improved readability and ease in maintenance of complex queries. The query can be divided into separate, simple, logical building blocks. These simple blocks can then be used to build more complex, interim CTEs until the final result set is generated. Using scalar subqueries (such as the (SELECT COUNT(1) FROM ...) examples we've looked at in this article) cannot be grouped or filtered directly in the containing query. Similarly, when using SQL Server 2005's ranking functions - ROW_NUMBER(), RANK(), DENSE_RANK(), and so on - the containing query cannot include a filter or grouping expression to return only a subset of the ranked results. For both of these instances, CTEs are quite handy. (For more on SQL Server 2005's ranking capabilities, be sure to read: Returning Ranked Results with Microsoft SQL Server 2005.) CTEs can also be used to recursively enumerate hierarchical data. We'll examine this next!

Recursive Common Table Expressions
Recursion is the process of defining a solution to a problem in terms of itself. For example, a teacher needs to sort a stack of tests alphabetically by the students' names. She could process the tests one at a time and, for each test, insert it into the appropriate spot to the left (called insertion sort), probably the way most people sort a hand of cards (at least that's the way I do it). However, depending on the distribution of the tests, the size of the work space, the number of tests to sort, and so on, it may be far more efficient to break down the problem into pieces. Rather than doing an insertion sort right off the bat, it might first make sense to divide the stack of papers in half, and then do an insertion sort on one half, an insertion sort on the second half, and then a merge of the two piles. Or perhaps it would make sense to divide the tests into four piles, or eight piles. (This approach is referred to as merge sort.)

With a recursive solution you will always have the following two pieces:

• The base case - what to do when you're done recursing. After dividing the tests into separate piles of say, eight elements per pile, the base case is to sort these piles via insertion sort.
• The recursive step - the action to perform that involves plugging the input "back into" the system. For merge sort, the recursive step is the division of one pile into two. Then into four. Then into eight, and so on, until the base case is reached.

For more on recursion, see Recursion, Why It's Cool.

Returning to CTEs... the Employees database table holds the corporate hierarchy within its rows. Imagine that good ol' Andrew Fuller has come back and insisted on a report that would list all persons in the company along with their position in the hierarchy. Since the Employees table can capture an arbitrary number of hierarchy levels, we need a recursive solution. Enter CTEs...

Like any recursive definition, a recursive Common Table Expression requires both a base case and the recursive step. In SQL parlance, this translates into two SQL queries - one that gets the "initial" data UNIONed with one that performs the recursion. For the Employees example, the base case is returning those employees that have no manager:

SELECT ... FROM Employees WHERE ReportsTo IS NULL

The recursion includes a query on the CTE itself. The following shows the CTE - with both the base case and recursive step - along with a SELECT query that returns the rows from the CTE:

WITH EmployeeHierarchy (EmployeeID, LastName, FirstName, ReportsTo, HierarchyLevel) AS ( -- Base case SELECT EmployeeID, LastName, FirstName, ReportsTo, 1 as HierarchyLevel FROM Employees WHERE ReportsTo IS NULL UNION ALL -- Recursive step SELECT e.EmployeeID, e.LastName, e.FirstName, e.ReportsTo, eh.HierarchyLevel + 1 AS HierarchyLevel FROM Employees e INNER JOIN EmployeeHierarchy eh ON e.ReportsTo = eh.EmployeeID ) SELECT * FROM EmployeeHierarchy ORDER BY HierarchyLevel, LastName, FirstName

The recursion occurs in the second query in the CTE by joining the results of Employees against the CTE itself (EmployeeHierarchy) where the employees' ReportsTo field matches up to the CTE's EmployeeID. Included in this query is the HierarchyLevel field, which returns 1 for the base case and one greater than the previous level for each recursive step down the hierarchy. As requested, this resultset clearly shows that Mr. Fuller is the alpha male in this organization. Furthermore, we can see that Steven, Laura, Nancy, Janet, and Margaret make up the second tier in the organizational hierarchy, while poor Anne, Robert, and Michael are down at the bottom:

Alternatives to Recursive Common Table Expressions

As we saw in this article, enumerating hierarchical data recursively can be accomplished via CTEs (for more on using recursive CTEs, don't forget to check out the official documentation - Recursive Queries Using Common Table Expressions). However, there are other options as well. One choice is to perform the recursion at the ASP/ASP.NET layer. That is, read in all employee information to a Recordset of DataSet in code, and then recurse there. My article Efficiently DisplayingParent-Child Data discusses this approach. If you need to perform the recursion in SQL, you can use recursive stored procedures, as discussed in The Zen of Recursion. If you are designing a data model that needs to support hierarchical data, your best bet is to bake in some lineage information directly into the table from the get-go. See SQL for Threaded Discussions and More Trees & Hierarchies in SQL for more information.

Conclusion
One of the many new features in SQL Server 2005 are Common Table Expressions (CTEs), which provide a more readable and usable approach to derived tables. Additionally, CTEs may be recursively defined, allowing a recursive entity to be enumerated without the need for recursive stored procedures. For more on the new features found in SQL Server 2005, be sure to also check out Returning Ranked Results with Microsoft SQL Server 2005 and TRY...CATCH in SQL Server 2005.