User-defined functions (UDFs) were introduced in SQL Server 2000 to provide the coding functionality similar to what is available in general-purpose programming languages. UDFs provide a level of flexibility and reusability that would be impossible to duplicate by using only stored procedures, triggers, and views. These functions, like those in other programming languages, provide reusable code modules that you can invoke with parameters, perform some kind of calculations or actions, and return various
kinds of data.
SQL Server supports three kinds of user-defined functions, each with its own specific features and benefits:
- Scalar functions are what most developers think of as user-defined functions, because they are most like the functions available in other programming languages and the built-in Transact-SQL (T-SQL) functions. A scalar function can take parameters, contain complex processing, and return a single value of almost any SQL data type.
- Inline table-valued functions are similar to views in that they each consist of a single SELECT statement and each returns a table object. However, inline table-valued functions can take parameters, whereas views cannot. Like views, you can use inline table-valued functions in statements that update, insert, or delete data, and you can join to them. An inline table-valued function is essentially a view with parameters.
- Multi-statement table-valued functions are similar to stored procedures, but they return an in-memory table object (a set of rows and columns that you can select from) instead of a static result set. This makes them a powerful view/stored procedure hybrid, because you can perform complex processing in them as in a stored procedure and select data from them as in a view. However, the tables that these functions return are not updatable.
TIP: Another way to categorize the types of functions available in SQL Server is
as scalar and table-valued, based on the type of data returned. With that
categorization, there are then two types of table-valued functions.
UDFs are not allowed to have side effects, which is a bit of a limitation on their use. This means that they are not allowed to make any lasting changes in the database, such as to make changes to data or to alter the schema of any object. But there are more subtle actions that cause changes and therefore are not allowed in UDFs. For example, the RAND function, which generates random numbers, retains state information about the last random number it generated, which changes the state of the function and requires SQL Server to store that state somewhere.
User-defined functions combine some of the best elements of stored procedures and views. Many tasks that you previously performed by using views and stored procedures you can now handle better with UDFs. You can invoke these functions from views, stored procedures, other user-defined functions, batches, and scripts, as well as from external applications.
NOTE: SQL Server 2000 did not permit nondeterministic functions in the body of a user-defined function. A deterministic function, like LEN, always returns the same result anytime it is called with the same argument values. A nondeterministic function, such as GETDATE(), always returns a different value on each invocation. However, in SQL Server 2005 and later, things have loosened up: GETDATE() is now permissible in UDFs.
Common Language Runtime UDFs
You can create each type of UDF in SQLCLR code as well as Transact-SQL. Transact-SQL is recommended for functions that rely on set-related data manipulation and SQLCLR functions are recommended for functions that require intensive computations or string manipulations, neither of which T-SQL is well suited for. In some cases, the .NET Framework contains base classes that supply functionality that would be difficult to duplicate in Transact-SQL, such as string processing that uses the RegEx class. However, if your functions involve fairly straightforward data manipulation, Transact-SQL is the best choice. This chapter will focus on T-SQL UDFs.
Don Kiely is a featured instructor on many of our SQL Server and Visual Studio courses. He is a nationally recognized author, instructor and consultant who travels the country sharing his expertise in SQL Server and security.