CREATE FUNCTION (Transact-SQL)

schema_name

Is the name of the schema to which the user-defined function belongs.

function_name

Is the name of the user-defined function. Function names must comply with the rules for identifiers and must be unique within the database and to its schema.

Note
Parentheses are required after the function name even if a parameter is not specified.

@parameter_name

Is a parameter in the user-defined function. One or more parameters can be declared.

A function can have a maximum of 2,100 parameters. The value of each declared parameter must be supplied by the user when the function is executed, unless a default for the parameter is defined.

Specify a parameter name by using an at sign (@) as the first character. The parameter name must comply with the rules for identifiers. Parameters are local to the function; the same parameter names can be used in other functions. Parameters can take the place only of constants; they cannot be used instead of table names, column names, or the names of other database objects.

Note
ANSI_WARNINGS is not honored when you pass parameters in a stored procedure, user-defined function, or when you declare and set variables in a batch statement. For example, if a variable is defined as char(3), and then set to a value larger than three characters, the data is truncated to the defined size and the INSERT or UPDATE statement succeeds.

[ type_schema_name. ] parameter_data_type

Is the parameter data type, and optionally the schema to which it belongs. For Transact-SQL functions, all data types, including CLR user-defined types and user-defined table types, are allowed except the timestamp data type. For CLR functions, all data types, including CLR user-defined types, are allowed except text, ntext, image, user-defined table types and timestamp data types. The nonscalar types, cursor and table, cannot be specified as a parameter data type in either Transact-SQL or CLR functions.

If type_schema_name is not specified, the Database Engine looks for the scalar_parameter_data_type in the following order:

• The schema that contains the names of SQL Server system data types.

• The default schema of the current user in the current database.

• The dbo schema in the current database.

[ =default ]

Is a default value for the parameter. If a default value is defined, the function can be executed without specifying a value for that parameter.

Note
Default parameter values can be specified for CLR functions except for the varchar(max) and varbinary(max) data types.

When a parameter of the function has a default value, the keyword DEFAULT must be specified when the function is called to retrieve the default value. This behavior is different from using parameters with default values in stored procedures in which omitting the parameter also implies the default value.

READONLY

Indicates that the parameter cannot be updated or modified within the definition of the function. If the parameter type is a user-defined table type, READONLY should be specified.

return_data_type

Is the return value of a scalar user-defined function. For Transact-SQL functions, all data types, including CLR user-defined types, are allowed except the timestamp data type. For CLR functions, all data types, including CLR user-defined types, are allowed except the text, ntext, image, and timestamp data types. The nonscalar types, cursor and table, cannot be specified as a return data type in either Transact-SQL or CLR functions. 

function_body

Specifies that a series of Transact-SQL statements, which together do not produce a side effect such as modifying a table, define the value of the function. function_body is used only in scalar functions and multistatement table-valued functions.

In scalar functions, function_body is a series of Transact-SQL statements that together evaluate to a scalar value.

In multistatement table-valued functions, function_body is a series of Transact-SQL statements that populate a TABLE return variable.

scalar_expression

Specifies the scalar value that the scalar function returns.

TABLE

Specifies that the return value of the table-valued function is a table. Only constants and @local_variables can be passed to table-valued functions.

In inline table-valued functions, the TABLE return value is defined through a single SELECT statement. Inline functions do not have associated return variables.

In multistatement table-valued functions, @return_variable is a TABLE variable, used to store and accumulate the rows that should be returned as the value of the function. @return_variable can be specified only for Transact-SQL functions and not for CLR functions.

select_stmt

Is the single SELECT statement that defines the return value of an inline table-valued function.

ORDER (<order_clause>)

Specifies the order in which results are being returned from the table-valued function. For more information, see the section, "Guidance on Using Sort Order," later in this topic.

EXTERNAL NAME <method_specifier> assembly_name.class_name.method_name

Specifies the method of an assembly to bind with the function. assembly_name must match an existing assembly in SQL Server in the current database with visibility on. class_name must be a valid SQL Server identifier and must exist as a class in the assembly. If the class has a namespace-qualified name that uses a period (.) to separate namespace parts, the class name must be delimited by using brackets ([ ]) or quotation marks (" "). method_name must be a valid SQL Server identifier and must exist as a static method in the specified class.

Note
By default, SQL Server cannot execute CLR code. You can create, modify, and drop database objects that reference common language runtime modules; however, you cannot execute these references in SQL Server until you enable the clr enabled option. To enable this option, use sp_configure.

<table_type_definition> ( { <column_definition> <column_constraint>   | <computed_column_definition> }   [ <table_constraint> ] [ ,...n ] )

Defines the table data type for a Transact-SQL function. The table declaration includes column definitions and column or table constraints. The table is always put in the primary filegroup.

< clr_table_type_definition > ( { column_namedata_type } [ ,...n ] )

Defines the table data types for a CLR function. The table declaration includes only column names and data types. The table is always put in the primary filegroup.

If a user-defined function is not created with the SCHEMABINDING clause, changes that are made to underlying objects can affect the definition of the function and produce unexpected results when it is invoked. We recommend that you implement one of the following methods to ensure that the function does not become outdated because of changes to its underlying objects:

• Specify the WITH SCHEMABINDING clause when you are creating the function. This ensures that the objects referenced in the function definition cannot be modified unless the function is also modified.

• Execute the sp_refreshsqlmodule stored procedure after modifying any object that is specified in the definition of the function.

If parameters are specified in a CLR function, they should be SQL Server types as defined previously for scalar_parameter_data_type. For information about comparing SQL Server system data types to CLR integration data types or .NET Framework common language runtime data types, see Mapping CLR Parameter Data.

For SQL Server to reference the correct method when it is overloaded in a class, the method indicated in <method_specifier> must have the following characteristics:

• Receive the same number of parameters as specified in [ ,...n ].

• Receive all the parameters by value, not by reference.

• Use parameter types that are compatible with those specified in the SQL Server function.

If the return data type of the CLR function specifies a table type (RETURNS TABLE), the return data type of the method in <method_specifier> should be of type IEnumerator or IEnumerable, and it is assumed that the interface is implemented by the creator of the function. Unlike Transact-SQL functions, CLR functions cannot include PRIMARY KEY, UNIQUE, or CHECK constraints in <table_type_definition>. The data types of columns specified in <table_type_definition> must match the types of the corresponding columns of the result set returned by the method in <method_specifier> at execution time. This type-checking is not performed at the time the function is created.

For more information about how to program CLR functions, see CLR User-Defined Functions.

Scalar-valued functions can be invoked where scalar expressions are used. This includes computed columns and CHECK constraint definitions. Scalar-valued functions can also be executed by using the EXECUTE statement. Scalar-valued functions must be invoked by using at least the two-part name of the function. For more information about multipart names, see Transact-SQL Syntax Conventions (Transact-SQL). Table-valued functions can be invoked where table expressions are allowed in the FROM clause of SELECT, INSERT, UPDATE, or DELETE statements. For more information, see Executing User-Defined Functions (Database Engine).

The following statements are valid in a function:

• Assignment statements.

• Control-of-Flow statements except TRY...CATCH statements.

• DECLARE statements defining local data variables and local cursors.

• SELECT statements that contain select lists with expressions that assign values to local variables.

• Cursor operations referencing local cursors that are declared, opened, closed, and deallocated in the function. Only FETCH statements that assign values to local variables using the INTO clause are allowed; FETCH statements that return data to the client are not allowed.

• INSERT, UPDATE, and DELETE statements modifying local table variables.

• EXECUTE statements calling extended stored procedures.

• For more information, see Creating User-Defined Functions (Database Engine).

Computed Column Interoperability

Calling Extended Stored Procedures from Functions

User-defined functions cannot be used to perform actions that modify the database state.

User-defined functions cannot contain an OUTPUT INTO clause that has a table as its target.

The following Service Broker statements cannot be included in the definition of a Transact-SQL user-defined function:

• BEGIN DIALOG CONVERSATION

• END CONVERSATION

• GET CONVERSATION GROUP

• MOVE CONVERSATION

• RECEIVE

• SEND

User-defined functions can be nested; that is, one user-defined function can call another. The nesting level is incremented when the called function starts execution, and decremented when the called function finishes execution. User-defined functions can be nested up to 32 levels. Exceeding the maximum levels of nesting causes the whole calling function chain to fail. Any reference to managed code from a Transact-SQL user-defined function counts as one level against the 32-level nesting limit. Methods invoked from within managed code do not count against this limit.

Using Sort Order in CLR Table-valued Functions

The following table lists the system catalog views that you can use to return metadata about user-defined functions.

USE AdventureWorks;
GO
SELECT definition, type 
FROM sys.sql_modules AS m
JOIN sys.objects AS o ON m.object_id = o.object_id 
    AND type IN ('FN', 'IF', 'TF');
GO

Requires CREATE FUNCTION permission in the database and ALTER permission on the schema in which the function is being created. If the function specifies a user-defined type, requires EXECUTE permission on the type.

A. Using a scalar-valued user-defined function that calculates the ISO week

USE AdventureWorks;
GO
IF OBJECT_ID (N'dbo.ISOweek', N'FN') IS NOT NULL
    DROP FUNCTION dbo.ISOweek;
GO
CREATE FUNCTION dbo.ISOweek (@DATE datetime)
RETURNS int
WITH EXECUTE AS CALLER
AS
BEGIN
     DECLARE @ISOweek int;
     SET @ISOweek= DATEPART(wk,@DATE)+1
          -DATEPART(wk,CAST(DATEPART(yy,@DATE) as CHAR(4))+'0104');
--Special cases: Jan 1-3 may belong to the previous year
     IF (@ISOweek=0) 
          SET @ISOweek=dbo.ISOweek(CAST(DATEPART(yy,@DATE)-1 
               AS CHAR(4))+'12'+ CAST(24+DATEPART(DAY,@DATE) AS CHAR(2)))+1;
--Special case: Dec 29-31 may belong to the next year
     IF ((DATEPART(mm,@DATE)=12) AND 
          ((DATEPART(dd,@DATE)-DATEPART(dw,@DATE))>= 28))
          SET @ISOweek=1;
     RETURN(@ISOweek);
END;
GO
SET DATEFIRST 1;
SELECT dbo.ISOweek(CONVERT(DATETIME,'12/26/2004',101)) AS 'ISO Week';

B. Creating an inline table-valued function

USE AdventureWorks;
GO
IF OBJECT_ID (N'Sales.ufn_SalesByStore', N'IF') IS NOT NULL
    DROP FUNCTION Sales.ufn_SalesByStore;
GO
CREATE FUNCTION Sales.ufn_SalesByStore (@storeid int)
RETURNS TABLE
AS
RETURN 
(
    SELECT P.ProductID, P.Name, SUM(SD.LineTotal) AS 'YTD Total'
    FROM Production.Product AS P 
      JOIN Sales.SalesOrderDetail AS SD ON SD.ProductID = P.ProductID
      JOIN Sales.SalesOrderHeader AS SH ON SH.SalesOrderID = SD.SalesOrderID
    WHERE SH.CustomerID = @storeid
    GROUP BY P.ProductID, P.Name
);
GO

SELECT * FROM Sales.ufn_SalesByStore (602);

C. Creating a multi-statement table-valued function

USE AdventureWorks;
GO
IF OBJECT_ID (N'dbo.ufn_FindReports', N'TF') IS NOT NULL
    DROP FUNCTION dbo.ufn_FindReports;
GO
CREATE FUNCTION dbo.ufn_FindReports (@InEmpID INTEGER)
RETURNS @retFindReports TABLE 
(
    EmployeeID int primary key NOT NULL,
    Name nvarchar(255) NOT NULL,
    Title nvarchar(50) NOT NULL,
    EmployeeLevel int NOT NULL,
    Sort nvarchar (255) NOT NULL
)
--Returns a result set that lists all the employees who report to the 
--specific employee directly or indirectly.*/
AS
BEGIN
   WITH DirectReports(Name, Title, EmployeeID, EmployeeLevel, Sort) AS
    (SELECT CONVERT(Varchar(255), c.FirstName + ' ' + c.LastName),
        e.Title,
        e.EmployeeID,
        1,
        CONVERT(Varchar(255), c.FirstName + ' ' + c.LastName)
     FROM HumanResources.Employee AS e
          JOIN Person.Contact AS c ON e.ContactID = c.ContactID 
     WHERE e.EmployeeID = @InEmpID
   UNION ALL
     SELECT CONVERT(Varchar(255), REPLICATE ('| ' , EmployeeLevel) +
        c.FirstName + ' ' + c.LastName),
        e.Title,
        e.EmployeeID,
        EmployeeLevel + 1,
        CONVERT (Varchar(255), RTRIM(Sort) + '| ' + FirstName + ' ' + 
                 LastName)
     FROM HumanResources.Employee as e
          JOIN Person.Contact AS c ON e.ContactID = c.ContactID
          JOIN DirectReports AS d ON e.ManagerID = d.EmployeeID
    )
-- copy the required columns to the result of the function 
   INSERT @retFindReports
   SELECT EmployeeID, Name, Title, EmployeeLevel, Sort
   FROM DirectReports 
   RETURN
END;
GO
-- Example invocation
SELECT EmployeeID, Name, Title, EmployeeLevel
FROM dbo.ufn_FindReports(109)
ORDER BY Sort;
GO

D. Creating a CLR function

DECLARE @SamplesPath nvarchar(1024);
-- You may have to modify the value of this variable if you have
-- installed the sample in a location other than the default location.
SELECT @SamplesPath = REPLACE(physical_name, 'Microsoft SQL Server\MSSQL10.MSSQLSERVER\MSSQL\DATA\master.mdf', 'Microsoft SQL Server\90\Samples\Engine\Programmability\CLR\') 
    FROM master.sys.database_files 
    WHERE name = 'master';

CREATE ASSEMBLY [SurrogateStringFunction]
FROM @SamplesPath + 'StringManipulate\CS\StringManipulate\bin\debug\SurrogateStringFunction.dll'
WITH PERMISSION_SET = EXTERNAL_ACCESS;
GO

CREATE FUNCTION [dbo].[len_s] (@str nvarchar(4000))
RETURNS bigint
AS EXTERNAL NAME [SurrogateStringFunction].[Microsoft.Samples.SqlServer.SurrogateStringFunction].[LenS];
GO

Reference

Other Resources