380 CHAPTER 11 ■ T-SQL ESSENTIALS UPPER() The final example is the reverse of the LOWER() function and changes all characters to uppercase. Try It Out: UPPER() 1. After the declared variable has been set, we then use the UPPER() function to change the value to uppercase. DECLARE @StringTest char(10) SET @StringTest = 'Robin ' SELECT UPPER(@StringTest) 2. And as you can see from Figure 11-36, Robin becomes ROBIN. Figure 11-36. Changing the case of a string to uppercase System Functions System functions are functions that provide extra functionality outside of the boundaries that can be defined as string, numeric, or date related. Three of these functions will be used extensively throughout our code, and therefore you should pay special attention to CASE, CAST, and ISNULL. CASE WHEN. . .THEN. . .ELSE. . .END The first function is when we wish to test a condition. WHEN that condition is true THEN we can do further processing, ELSE if it is false, then we can do something else. What happens in the WHEN section and the THEN section can range from another CASE statement to providing a value that sets a column or a variable. The CASE WHEN statement can be used to return a value or, if on the right-hand side of an equality statement, to set a value. Both of these scenarios are covered in the following examples. Try It Out: CASE 1. The example will use a CASE statement to add up customers’ TransactionDetails.Transactions for the month of August. If the TransactionType is 0, then this is a Debit; if it is a 1, then it is a Credit. By using the SUM aggregation, we can add up the amounts. Combine this with a GROUP BY where the TransactionDetails. Transactions are split between Credit and Debit, and we get two rows in the results set: one for debits and one for credits. SET QUOTED_IDENTIFIER OFF SELECT CustomerId, CASE WHEN CreditType = 0 THEN "Debits" ELSE "Credits" END Dewson_958-7.book Page 380 Monday, June 30, 2008 3:01 PM CHAPTER 11 ■ T-SQL ESSENTIALS 381 AS TranType,SUM(Amount) FROM TransactionDetails.Transactions t JOIN TransactionDetails.TransactionTypes tt ON tt.TransActionTypeId = t.TransactionType WHERE t.DateEntered BETWEEN '1 Aug 2008' AND '31 Aug 2008' GROUP BY CustomerId,CreditType 2. When the code is run, you should see the results shown in Figure 11-37. Figure 11-37. Decisions within a string CAST()/CONVERT() These are two functions used to convert from one data type to another. The main difference between them is that CAST() is ANSI SQL–92 compliant, but CONVERT() has more functionality. The syntax for CAST() is CAST(variable_or_column AS datatype) This is opposed to the syntax for CONVERT(), which is CONVERT(datatype,variable_or_column) Not all data types can be converted between each other, such as converting a datetime to a text data type, and some conversions need neither a CAST() nor a CONVERT(). There is a grid in Books Online that provides the necessary information. If you wish to CAST() from numeric to decimal or vice versa, then you need to use CAST(); other- wise, you will lose precision. Try It Out: CAST()/CONVERT() 1. The first example will use CAST to move a number to a char(10). DECLARE @Cast int SET @Cast = 1234 SELECT CAST(@Cast as char(10)) + '-End' 2. Executing this code results in a left-filled character variable, as shown in Figure 11-38. Figure 11-38. Changing the data type of a value Dewson_958-7.book Page 381 Monday, June 30, 2008 3:01 PM 382 CHAPTER 11 ■ T-SQL ESSENTIALS 3. The second example completes the same conversion, but this time we use the CONVERT() function. DECLARE @Convert int SET @Convert = 5678 SELECT CONVERT(char(10),@Convert) + '-End' 4. As you can see from Figure 11-39, the only change is the value output. Figure 11-39. Changing the data type of a value, using the non-ANSI standard ISDATE() Although ISDATE() is a function that works with dates and times, this system function takes a value in a column or a variable and confirms whether it contains a valid date or time. The value returned is 0, or false, for an invalid date, or 1 for true if the date is okay. The formatting of the date for testing within the ISDATE() function has to be in the same regional format as you have set with SET DATEFORMAT or SET LANGUAGE. If you are testing in a European format but have your database set to US format, then you will get a false value returned. Try It Out: ISDATE() 1. The first example demonstrates where a date is invalid. There are only 30 days in September. DECLARE @IsDate char(15) SET @IsDate = '31 Sep 2008' SELECT ISDATE(@IsDate) 2. Execute the code, and you should get the results shown in Figure 11-40. Figure 11-40. Testing if a value is a date 3. Our second example is a valid date. DECLARE @IsDate char(15) SET @IsDate = '30 Sep 2008' SELECT ISDATE(@IsDate) Dewson_958-7.book Page 382 Monday, June 30, 2008 3:01 PM CHAPTER 11 ■ T-SQL ESSENTIALS 383 4. This time when you run the code, you see a value of 1, as shown in Figure 11-41, denoting a valid entry. Figure 11-41. Showing that a value is a date ISNULL() Many times so far, you have seen NULL values within a column of returned data. As a value, NULL is very useful, as you have seen. However, you may wish to test whether a column contains a NULL or not. If there were a value, you would retain it, but if there were a NULL, you would convert it to a value. This function could be used to cover a NULL value in an aggregation, for example. The syntax is ISNULL(value_to_test,new_value) where the first parameter is the column or variable to test if there is a NULL value, and the second option defines what to change the value to if there is a NULL value. This change only occurs in the results and doesn’t change the underlying data that the value came from. Try It Out: ISNULL() 1. In this example, we define a char() variable of ten characters in length and then set the value explicitly to NULL. The example will also work without the second line of code, which is simply there for clarity. The third line tests the variable, and as it is NULL, it changes it to a date. Note, though, that a date is more than ten characters, so the value is truncated. DECLARE @IsNull char(10) SET @IsNull = NULL SELECT ISNULL(@IsNull,GETDATE()) 2. As expected, when you execute the code, you get the first ten characters of the relevant date, as shown in Figure 11-42. Figure 11-42. Changing the NULL to a value if the value is a NULL ISNUMERIC() This final system function tests the value within a column or variable and ascertains whether it is numeric or not. The value returned is 0, or false, for an invalid number, or 1 for true if the test is okay and can convert to a numeric. Dewson_958-7.book Page 383 Monday, June 30, 2008 3:01 PM 384 CHAPTER 11 ■ T-SQL ESSENTIALS ■Note Currency symbols such as £ and $ will also return 1 for a valid numeric value. Try It Out: ISNUMERIC() 1. Our first example to demonstrate ISNUMERIC() defines a character variable and contains alphabetic values. This test fails, as shown in Figure 11-43. DECLARE @IsNum char(10) SET @IsNum = 'Robin ' SELECT ISNUMERIC(@IsNum) Figure 11-43. Checking whether a value is a number and finding out it is not 2. This second example places numbers and spaces into a char field. The ISNUMERIC() test ignores the spaces, provided that there are no further alphanumeric characters. DECLARE @IsNum char(10) SET @IsNum = '1234 ' SELECT ISNUMERIC(@IsNum) Figure 11-44 shows the results of running this code. Figure 11-44. Finding out a value is numeric RAISERROR Before we look at handling errors, you need to be aware of what an error is, how it is generated, the information it generates, and how to generate your own errors when something is wrong. The T-SQL command RAISERROR allows us as developers to have the ability to produce our own SQL Server error messages when running queries or stored procedures. We are not tied to just using error messages that come with SQL Server; we can set up our own messages and our own level of severity for those messages. It is also possible to determine whether the message is recorded in the Windows error log or not. However, whether we wish to use our own error message or a system error message, we can still generate an error message from SQL Server as if SQL Server itself raised it. Enterprise environments typically experience the same errors on repeated occasions, since they employ SQL Server in very specific ways depending on their business model. With this in mind, attention to employing RAISERROR can have big benefits by providing more meaningful feedback as well as suggested solutions for users. Dewson_958-7.book Page 384 Monday, June 30, 2008 3:01 PM CHAPTER 11 ■ T-SQL ESSENTIALS 385 By using RAISERROR, the whole SQL Server system acts as if SQL Server raised the error, as you have seen within this book. RAISERROR can be used in one of two ways; looking at the syntax will make this clear. RAISERROR ({msg_id|msg_str} {,severity,state} [,argument [ , n ] ]) [WITH option [ , n ]] You can either use a specific msg_id or provide an actual output string, msg_str, either as a literal or a local variable defined as string-based, containing the error message that will be recorded. The msg_id references system and user-defined messages that already exist within the SQL Server error messages table. When specifying a text message in the first parameter of the RAISERROR function instead of a message ID, you may find that this is easier to write than creating a new message: RAISERROR('You made an error', 10, 1) The next two parameters in the RAISERROR syntax are numerical and relate to how severe the error is and information about how the error was invoked. Severity levels range from 1 at the innoc- uous end to 25 at the fatal end. Severity levels of 2 to 14 are generally informational. Severity level 15 is for warnings, and levels 16 or higher represent errors. Severity levels from 20 to 25 are considered fatal, and require the WITH LOG option, which means that the error is logged in the Windows Application Event log and the SQL Error log and the connection terminated; quite simply, the stored procedure stops executing. The connection referred to here is the connection within Query Editor, or the connection made by an application using a data access method like ADO.NET. Only for a most extreme error would we set the severity to this level; in most cases, we would use a number between 1 and 18. The last parameter within the function specifies state. Use a 1 here for most implementations, although the legitimate range is from 1 to 127. You may use this to indicate which error was thrown by providing a different state for each RAISERROR function in your stored procedure. SQL Server does not act on any legitimate state value, but the parameter is required. A msg_str can define parameters within the text. By placing the value, either statically or via a variable, after the last parameter that you define, msg_str replaces the message parameter with that value. This is demonstrated in an upcoming example. If you do wish to add a parameter to a message string, you have to define a conversion specification. The format is % [[flag] [width] [. precision] [{h | l}]] type The options are as follows: • flag: A code that determines justification and spacing of the value entered: • - (minus): Left-justify the value. • + (plus): The value shows a + or a – sign. • 0: Prefix the output with zeros. • #: Preface any nonzero with a 0, 0x, or 0X, depending on the formatting. • (blank): Prefix with blanks. • width: The minimum width of the output. • precision: The maximum number of characters used from the argument. Dewson_958-7.book Page 385 Monday, June 30, 2008 3:01 PM 386 CHAPTER 11 ■ T-SQL ESSENTIALS • h: Character types: • d or i: Signed integer. • o: Unsigned octal. • s: String. • u: Unsigned integer. • x or X: Unsigned hex. To place a parameter within a message string where the parameter needs to be inserted, you would define this by a % sign followed by one of the following options: d or i for a signed integer, p for a pointer, s for a string, u for an unsigned integer, x or X for an unsigned hexadecimal, and o for an unsigned octal. Note that float, double, and single are not supported as parameter types for messages. You will see this in action in the upcoming examples. Finally, there are three options that could be placed at the end of the RAISERROR message. These are the WITH options: • LOG places the error message within the Windows error log. • NOWAIT sends the error directly to the client. • SETERROR resets the error number to 50000 within the message string only. When using any of these last WITH options, do take the greatest of care, as their misuse can create more problems than they solve. For example, you may unnecessarily use LOG a great deal, filling up the Windows Application Event log and the SQL Error log, which leads to further problems. There is a system stored procedure, sp_addmessage, that can create a new global error message that can be used by RAISERROR by defining the @msgnum. The syntax for adding a message is sp_addmessage [@msgnum =]msg_id, [@severity = ] severity , [ @msgtext = ] 'msg' [ , [ @lang = ] 'language' ] [ , [ @with_log = ] 'with_log' ] [ , [ @replace = ] 'replace' ] The parameters into this system stored procedure are as follows: • @msgnum: The number of the message is typically greater than 50000. • @severity: Same as the preceding, in a range of 1 to 25. • @lang: Use this if you need to define the language of the error message. Normally this is left empty. • @with_log: Set to 'TRUE' if you wish to write a message to the Windows error log. • @replace: Set to 'replace' if you are replacing an existing message and updating any of the preceding values with new settings. ■Note Any message added will be specific for that database rather than the server. It is time to move to an example that will set up an error message that will be used to say a customer is overdrawn. Dewson_958-7.book Page 386 Monday, June 30, 2008 3:01 PM CHAPTER 11 ■ T-SQL ESSENTIALS 387 Try It Out: RAISERROR 1. First of all, we want to add a new user-defined error message. To do this, we will use sp_addmessage. We can now add any new SQL Server message that we wish. Any user-defined error message must be greater than 50000, so the first error message would normally be 50001. sp_addmessage @msgnum=50001,@severity=1, @msgtext='Customer is overdrawn' 2. We can then perform a RAISERROR to see the message displayed. Notice that we have to define the severity again. This is mandatory, but would be better if it was optional, and then you could always default to the severity defined. RAISERROR (50001,1,1) 3. When this is executed, we will see the following output: Customer is overdrawn Msg 50001, Level 1, State 1 4. This is not the friendliest of messages, as it would be better to perhaps give out the customer number as well. We can do this via a parameter. In the code that follows, we replace the message just added and now include a parameter where we are formatting with flag 0, which means we are prefixing the output with zeros; then we include the number 10, which is the precision, so that means the number will be ten digits; and finally we indicate the message will be unsigned using the option u. sp_addmessage @msgnum =50001,@severity=1, @msgtext='Customer is overdrawn. CustomerId= %010u',@replace='replace' 5. We can then change the RAISERROR so that we add on another parameter. We are hard coding the customer number as customer number 243, but we could use a local variable. RAISERROR (50001,1,1,243) 6. Executing the code now produces output that is much better and more informative for debugging, if required. Customer is overdrawn. CustomerId= 0000000243 Msg 50001, Level 1, State 1 Now that you know how you can raise your own errors if scenarios crop up that need them, we can take a look at how SQL Server can deal with errors. We do come back to RAISERROR when looking at these two options next. Error Handling When working with T-SQL, it is important to have some sort of error handling to cater to those times when something goes wrong. Errors can be of different varieties; for example, you might expect at least one row of data to be returned from a query, and then you receive no rows. However, what we are discussing here is when SQL Server informs us there is something more drastically wrong. We have seen some errors throughout the book, and even in this chapter. There are two methods of error catching we can employ in such cases. The first uses a system variable, @@ERROR. Dewson_958-7.book Page 387 Monday, June 30, 2008 3:01 PM 388 CHAPTER 11 ■ T-SQL ESSENTIALS @@ERROR This is the most basic of error handling. It has served SQL Server developers well over the years, but it can be cumbersome. When an error occurs, such as you have seen as we have gone through the book creating and manipulating objects, a global variable, @@ERROR, would have been populated with the SQL Server error message number. Similarly, if you try to do something with a set of data that is invalid, such as dividing a number by zero or exceeding the number of digits allowed in a numeric data type, then SQL Server will populate this variable for you to inspect. The downside is that the @@ERROR variable setting only lasts for the next statement following the line of code that has been executed; therefore, when you think there might be problems, you need to either pass the data to a local variable or inspect it straight away. The first example demonstrates this. Try It Out: Using @@ERROR 1. This example tries to divide 100 by zero, which is an error. We then list out the error number, and then again list out the error number. Enter the following code and execute it: SELECT 100/0 SELECT @@ERROR SELECT @@ERROR 2. It is necessary in this instance to check both the Results and Messages tab. The first tab is the Messages tab, which shows you the error that encountered. As expected, we see the Divide by zero error encountered message. Msg 8134, Level 16, State 1, Line 1 Divide by zero error encountered. (1 row(s) affected) (1 row(s) affected) 3. Moving to the Results tab, you should see three result sets, as shown in Figure 11-45. The first, showing no infor- mation, would be where SQL Server would have put the division results, had it succeeded. The second result set is the number from the first SELECT @@ERROR. Notice the number corresponds to the msg number found in the Messages tab. The third result set shows a value of 0. This is because the first SELECT @@ERROR worked successfully and therefore set the system variable to 0. This demonstrates the lifetime of the value within @@ERROR. Figure 11-45. Showing @@ERROR in multiple statements Dewson_958-7.book Page 388 Monday, June 30, 2008 3:01 PM CHAPTER 11 ■ T-SQL ESSENTIALS 389 4. When we use the RAISERROR function, it also sets the @@ERROR variable, as we can see in the following code. However, the value will be set to 0 using our preceding example. This is because the severity level was below 11. RAISERROR (50001,1,1,243) SELECT @@ERROR 5. When the code is executed, you can see that @@ERROR is set to 0, as shown in Figure 11-46. Figure 11-46. When severity is too low to set @@ERROR 6. By changing the severity to 11 or above, the @@ERROR setting will now be set to the message number within the RAISERROR. RAISERROR (50001,11,1,243) SELECT @@ERROR 7. The preceding code produces the same message as seen within our RAISERROR example, but as you can see in Figure 11-47, the error number setting now reflects that value placed in the msgnum parameter. Figure 11-47. With a higher severity, the message number is set. Although a useful tool, it would be better to use the next error-handling routine to be demonstrated, TRY CATCH. TRY. . .CATCH It can be said that no matter what, any piece of code has the ability to fail and generate some sort of error. For the vast majority of this code, you will want to trap any error that occurs, check what the error is, and deal with it as best you can. As you saw previously, this could be done one statement at a time using @@ERROR to test for any error code. A new and improved functionality exists whereby a set of statements can try and execute, and if any statement has an error, it will be caught. This is known as a TRY CATCH block. Surrounding code with the ability to try and execute a slice of code and to catch any errors and try to deal with them has been around for quite a number of years in languages such as C++. Gladly, we now see this within SQL Server. The syntax is pretty straightforward. There are two “blocks” of code. The first block, BEGIN TRY, is where there is one or more T-SQL statements that you wish to try and run. If any of statements have an error, then no further processing within that block will execute, and processing will switch to the second block, BEGIN CATCH. Dewson_958-7.book Page 389 Monday, June 30, 2008 3:01 PM [...]... expect PowerShell to have a similar level of functional richness as VBA, PowerShell does provide the ability to produce featurerich tools that allow you to administer servers and improve the functionality within SQL Server agent-based jobs, similar to the maintenance plan you saw in Chapter 7 PowerShell uses SQL Server 2008 client connectivity to attach to a database When installing SQL Server in Chapter... selected to install The client connectivity components allow PowerShell to connect to a SQL Server installation from SQL Server 2000 onward, so you can leverage the functionality within PowerShell throughout your organization However, SQL Server 2005 does require Service Pack 2, and SQL Server 2000 requires Service Pack 4 The aim of this section is to introduce you to PowerShell, demonstrate how to find... the file a sql suffix Create a file called invoke-cmd-ex .sql with the following code within it: SELECT CustomerFirstName + ' ' + CustomerLastName FROM CustomerDetails.Customers SELECT 'Query2' SELECT CustomerFirstName + ' ' + CustomerLastName FROM CustomerDetails.Customers WHERE customerid > 1 DECLARE @return_value int, @ClearedBalance money, @UnclearedBalance money EXEC @return_value = [CustomerDetails].[apf_CustBalances]... invoke-sqlcmd -query "SELECT CustomerFirstName + ' ' + CustomerLastName FROM CustomerDetails.Customers" -serverinstance "FAT-BELLY" 8 However, what if you want to run a number of commands or you wish to run a command repetitively? It is possible to create a file that you can use as an input option with one or more SQL Server commands The suffix of the file can be anything, although it is usual to give... that for the transaction you are on in the main query However, you also need to ensure you are doing this for the same customer as the SELECT used previously You need to make a join from the main table to the subquery joining on the customer ID Without this, you could be selecting any minimum date entered from any customer SELECT MIN(DateEntered) FROM TransactionDetails.Transactions t2 WHERE t2.CustomerId... have a loose end from a previous chapter that we need to tidy up In Chapter 10, you built a scalar function to calculate interest between two dates using a rate It was called the TransactionDetails fn_IntCalc customer When I demonstrated this, you had to pass in the specific information; however, by using a subquery, it is possible to use this function to calculate interest for customers based on their... t2.CustomerId = t1.CustomerId AND t2.DateEntered> t1.DateEntered) as 'To Date', 3 Finally, you can put the same query into the call to the interest calculation function The final part is where you are just looking for the transactions for customer ID 1 TransactionDetails.fn_IntCalc (10, t1.Amount,t1.DateEntered, (SELECT MIN(DateEntered) FROM TransactionDetails.Transactions t2 WHERE t2.CustomerId = t1.CustomerId AND... alias of so From this table, you then instruct SQL Server to PIVOT the columns while completing an aggregation on a specific column—in our case, a SUM of the LineTotal column from the table value expression It is also at this point you define the columns to create via the FOR statement—in our case, a column for each product of the three products This is the equivalent to using GROUP BY for the aggregation,... deploying a NET assembly onto a server For example, not many ISPs allow you to deploy a NET-based assembly onto a shared server Dewson_958-7.book Page 413 Monday, June 30, 2008 3:01 PM C HA PTER 12 ■ A DVAN CED T -SQL PowerShell is also a scripting language, so for those of you who have used Visual Basic for Applications (VBA) within Excel or other products, you can use VBA to produce more feature-rich... available within PowerShell, and show you how to invoke SQL commands PowerShell will become a major part of SQL Server, and this section will give you a flavor of what could be possible You will see how to instantiate PowerShell from SQL Server Management Studio (SMSS) You could also use a command prompt and enter the command for powershell.exe, but you would need to load the snapins manually One example . found.', 16, 1) END RESTORE VERIFYONLY FROM DISK = 'C:Program FilesMicrosoft SQL Server MSSQL .10. MSSQLSERVERMSSQLBackup ApressFinancialApressFinancial_backup _20080 8061136.bak' WITH. ideal for producing a list of values to search for, or for producing a value from another table to set a column or a variable with. It is also possible to create a transient table of data to use. developers to have the ability to produce our own SQL Server error messages when running queries or stored procedures. We are not tied to just using error messages that come with SQL Server; we