Microsoft is releasing a new version of SQL Server after a gap of five years. This version (SQL Server 2005) introduces a horde of new attention-grabbing features like native XML storage and querying (using XQuery), CLR integration, SQL Web Services, query notifications, and the Service Broker. Apart from these high-profile enhancements, SQL Server 2005 also comes with improved T-SQL support.

Overview of T-SQL Enhancements
Most of SQL Server 2005's T-SQL enhancements focus on offering greater expressiveness to queries. We will look at each of these individual enhancements in detail. The list of enhancements includes:

• Relational operators
• TOP
• Recursive queries
• XML showplan
• Snapshot isolation level
• Data types
• DDL triggers
• DML with output
• Exception handling

The PIVOT operator is very similar to the TRANSFORM operator that Microsoft Access provides. It is used to rotate rows into columns. The best way to learn about PIVOT is through an example (see Figure 1).

This table lists the sales of a Toyota dealership. Notice that some models have more than one row for a particular year. Now we apply the following SELECT query (which uses the PIVOT operator) to the SalesSummary table. The result we get is shown in Figure 2.

SELECT *
FROM [SalesSummary]
PIVOT (SUM ([UnitsSold]) FOR [Year] IN ([2003], [2004])) AS [T]

The table was PIVOTed on the year column. Notice how the unique year values have taken the shape of columns and the corresponding number of units sold have been summed up.

The UNPIVOT operator does the opposite of what PIVOT does - it rotates columns into rows. Let's say we want to rotate the table in Figure 2 (let's call it Year WiseSalesSummary), so that the number of units sold is displayed for each year. All we have to do is apply the following SELECT query. The result of the UNPIVOT operation is shown in Figure 3.

SELECT *
FROM [YearWiseSalesSummary]
UNPIVOT (UnitsSold FOR [Year] in ([2003], [2004])) AS T

The APPLY is another new operator available in SQL Server 2005. When used in the FROM clause, it invokes a table-valued UDF for each row of a table. The UDF can optionally use the columns of the table as arguments.

Let's say we have a table called StudentScores that has the Math, Physics, and Chemistry scores of a bunch of students (see Figure 4).

Now, create the Calculate-Grade function, which takes the subject scores of each of the student one at a time and returns the grade of the student.

CREATE FUNCTION [CalculateGrade] 
(@math AS INT, @physics AS INT, @chemistry AS INT)
RETURNS  @GradeTable TABLE ([Grade] CHAR (1))
AS
BEGIN
	DECLARE @average FLOAT
	SET @average = (@math + @physics + @chemistry) / 3
	IF @average >= 90
		INSERT INTO @GradeTable VALUES ('A')
	ELSE IF @average >=80
		INSERT INTO @GradeTable VALUES ('B')
	ELSE IF @average >=70
		INSERT INTO @GradeTable VALUES ('C')
	ELSE
		INSERT INTO @GradeTable VALUES ('F')
	RETURN
END

The following query uses the APPLY relational operator to apply the CalculateGrade function on every row of the StudentScores table. We also pass the Math, Physics, and Chemistry scores to the CalculateGrade function.

SELECT [Name], I.Grade
FROM [StudentScores] AS O
CROSS APPLY CalculateGrade ([Math], [Physics], [Chemistry]) AS I

Figure 5 is an illustrative example of the APPLY operator. There are several ways (many of them better than using the APPLY operator) to achieve the same result.

TOP
In SQL Server, a SELECT query can use the TOP option to specify the number of rows that should be returned. The argument to TOP was a constant value of type BIGINT. The TOP option can also be qualified with the PERCENT keyword to specify that a percentage of the total number of rows should be returned by a SELECT query. The argument in this case was a constant value of type FLOAT.

In SQL Server 2005, an argument can be specified using an expression or a query, the only condition being that both should result in a value that is of the correct type.

For example, the following query returns the latest 10 orders from the Orders table in the Northwind database.

DECLARE @NO_OF_ROWS BIGINT
SET @NO_OF_ROWS = 10

SELECT TOP (@NO_OF_ROWS)
*
FROM [Orders]
ORDER BY [OrderDate] DESC

In SQL Server 2005, the TOP keyword can also be used with INSERT, UPDATE, and DELETE commands. Using TOP to do batch inserts, updates, or delete is more optimal than using the older SET ROWCOUNT method.

An example of using TOP with the DELETE operator would be the following query, which purges the oldest 10,000 rows from TempTable.

DELETE TOP (10000)
FROM [TempTable]
ORDER BY [TimeStamp]

Recursive Queries
SQL Server 2005 introduces Common Table Expressions (CTEs) that allow you to write named table expressions that persist for the duration of a query. The functionality of CTEs is somewhere between derived tables and views.

A CTE can be defined by using the WITH clause, as shown in the following code.

WITH CustomerSales ([CustomerId, [TotalSales])
AS
(
SELECT
[CustomerId], SUM ([UnitPrice] * [Quantity] * (1 - [Discount]))
AS [TotalSales]
FROM [Orders] AS [O] INNER JOIN
[Order Details] AS [OD]
ON [O].[OrderId] = [OD].[OrderId]

GROUP BY [CustomerId]
)
SELECT * FROM [CustomerSales]

This query creates a non-recursive CTE called CustomerSales that has two columns: CustomerId and TotalSales. The CTE gets all unique customer ids and the total sales made to them.

The primary purpose of having CTEs is to support recursion. Recursive CTEs improve readability and manageability of complex SQL statements. Recursive CTEs have the ability to traverse recursive hierarchies in a single query. A typical scenario where one would use a recursive CTE would be when a table has a self-referential constraint.

The recursive form of a CTE is as shown:

<Non-recursive SELECT>
UNION ALL
<SELECT referencing the CTE>

The recursion terminates when the second SELECT block produces an empty result.

The following recursive CTE prints out the name of all employees and their managers found in the Northwind database. The results are shown in Figure 6.

WITH EmployeeManagerCTE ([EmployeeId], [EmployeeName], [ManagerName])
AS
(
	SELECT 
		[EmployeeId], 
		[LastName] + ', ' + [FirstName], 
		[LastName] + ', ' + [FirstName] 
	FROM employees 
	WHERE [ReportsTo] IS NULL

	UNION ALL

	SELECT 
		[E1].[EmployeeId], 
		[E1].[LastName] + ', ' + [E1].[FirstName], 
		[E2].[EmployeeName]
	FROM [Employees] [E1] INNER JOIN 
	[EmployeeManagerCTE] [E2]
	ON [E1].[ReportsTo] = [E2].[EmployeeId]

)
SELECT * FROM [EmployeeManagerCTE]

XML Showplan
SQL Server 2005 has the facility to output a query plan as XML that uses a documented XML schema and carries more information than SQL Server 2000 textual or graphical formats.

The XML query plan output lends itself to various kinds of advanced processing with XQuery, XSLT, DOM, and the CLR.

Snapshot Isolation Level
SQL Server 2000 supports four isolation levels: read uncommitted, read committed, repeatable read, and serializable. SQL Server 2005 introduces a new snapshot isolation level. The snapshot isolation level is based on optimistic concurrency. It holds onto a snapshot of committed data from when the transaction started, allowing non-blocking reads to take place.

Data Types
SQL Server 2005 introduces a few new data types: XML, VARCHAR (MAX), NVARCHAR (MAX), and VARBINARY (MAX).

The XML data type is a first-class data type in SQL Server 2005 and can be used to define columns, variables, and parameters for stored procedures and user-defined functions. The XML data type can be optionally constrained by an XML schema. SQL Server 2005 also provides built-in XQuery support to query XML data.

The VARCHAR (MAX), NVARCHAR (MAX), and VARBINARY (MAX) data types can store up to 2GB of data and provide alternatives to the existing TEXT, NTEXT, and IMAGE data types respectively.

DDL Triggers
SQL Server 2005 has extended traditional triggers and enabled them to be fired for DDL operations like CREATE, ALTER, and DROP. These triggers are created, altered, and dropped using similar Transact-SQL syntax as that for standard triggers.

DDL triggers apply to all commands of a single type across a database or server and they execute only after completion of a DDL statement (although they can roll back the transaction to before the execution of the DDL statement that caused their triggering). DDL triggers cannot be INSTEAD OF triggers.

DDL triggers help to enforce development rules and standards for objects in a database; protect from accidental drops; help in object check-in/checkout, source versioning, and log management activities.

An example of a DDL trigger that prevents the dropping of a table is as shown:

CREATE TRIGGER OnDropTable ON DATABASE FOR DROP_TABLE
AS
    RAISERROR ('No drops allowed', 10, 1)
ROLLBACK

DML with Output
SQL Server 2005 introduces the OUTPUT clause that returns inserted, updated, or deleted rows as part of a DML operation. The OUPUT clause can be used with INSERT, DELETE, and UPDATE statements. The "DELETED" and "INSERTED" pseudo-tables can be used to get the pre- and post-operation values.

An example usage is as follows:

UPDATE [Jobs]
SET [Status] = 'Done'
OUTPUT DELETED.*, INSERTED.*
WHERE [Status] = 'In progress'

The DELETED values are not available for INSERT operations while the INSERTED values are not available for the DELETE operations. Both the INSERTED and the DELETED values are available for UPDATE operations.

Exception Handling
Exception handling can now be done in SQL Server 2005 by using the very well-known TRY/CATCH paradigm. All errors that set the @@error variable will now throw exceptions. The syntax of using exception handling is as follows:

BEGIN TRY
    -- SQL statements
END TRY
BEGIN CATCH
    -- SQL statements
END CATCH

TRY/CATCH blocks can also be nested. Error details are available inside the CATCH block through a set of functions - error_number (), error_message (), error_severity (), and error_state (). RAISERROR can be used to re-throw and exception from within a CATCH block.

Conclusion
SQL Server 2005 enhances T-SQL on a number of fronts - it introduces new and important data types, brings more power and expressiveness to SQL querying, provides more procedural capabilities, and improves parallelism.