CREATE FUNCTION (SQL table)

This CREATE FUNCTION (SQL table) statement creates an SQL table function at the current server. The function returns a single result table.

Invocation

This statement can be embedded in an application program or issued interactively. It is an executable statement that can be dynamically prepared.

Authorization

The privileges held by the authorization ID of the statement must include at least one of the following:

The privilege to create in the schema. For more information, see Privileges necessary to create in a schema.
Database administrator authority

The privileges held by the authorization id of the statement must include at least one of the following:

For the SYSFUNCS catalog view and SYSPARMS catalog table:
- The INSERT privilege on the table, and
- The system authority *EXECUTE on library QSYS2
Database administrator authority

The privileges held by the authorization ID of the statement must also include at least one of the following:

The following system authorities:
- *USE to the Create Service Program (CRTSRVPGM) command or
Database administrator authority

If a distinct type is referenced, the privileges held by the authorization ID of the statement must include at least one of the following:

For each distinct type identified in the statement:
- The USAGE privilege on the distinct type, and
- The system authority *EXECUTE on the library containing the distinct type
Database administrator authority

If the SECURED attribute is specified, or the function is secure and OR REPLACE is specified:

The authorization ID of the statement must have security administrator authority. See Administrative authority.

To replace an existing function, the privileges held by the authorization ID of the statement must include at least one of the following:

The following system authorities:
- The system authority of *OBJMGT on the service program object associated with the function
- All authorities needed to DROP the function
- The system authority *READ to the SYSFUNCS catalog view and SYSPARMS catalog table
Database administrator authority

For information about the system authorities corresponding to SQL privileges, see Corresponding System Authorities When Checking Privileges to a Table or View and Corresponding System Authorities When Checking Privileges to a Distinct Type.

Syntax

function-definition

parameter-declaration

data-type1, data-type2

built-in-type

ccsid-clause

normalize-clause

default-clause

option-list

Notes:

¹ This clause and the clauses that follow in the option-list can be specified in any order. Each clause can be specified at most once.

SQL-routine-body

Description

OR REPLACE

Specifies to replace the definition for the function if one exists at the current server. The existing definition is effectively dropped before the new definition is replaced in the catalog with the exception that privileges that were granted on the function are not affected. This option is ignored if a definition for the function does not exist at the current server. To replace an existing function, the specific-name and function-name of the new definition must be the same as the specific-name and function-name of the old definition, or the signature of the new definition must match the signature of the old definition. Otherwise, a new function is created.

function-name

Names the user-defined function. The combination of name, schema name, the number of parameters, and the data type of each parameter (without regard for any length, precision, scale, or CCSID attributes of the data type) must not identify a user-defined function that exists at the current server unless OR REPLACE is specified.

For SQL naming, the function will be created in the schema specified by the implicit or explicit qualifier.

For system naming, the function will be created in the schema that is specified by the qualifier. If no qualifier is specified:

If the value of the CURRENT SCHEMA special register is *LIBL, the function will be created in the current library (*CURLIB).
Otherwise, the function will be created in the current schema.

In general, more than one function can have the same name if the function signature of each function is unique.

Certain function names are reserved for system use. For more information see Choosing the schema and function name in CREATE FUNCTION.

(parameter-declaration,...)

Specifies the number of input parameters of the function and the data type of each parameter. Each parameter-declaration specifies an input parameter for the function. A maximum of Start of change

2000

parameters can be specified. A function can have zero or more input parameters. There must be one entry in the list for each parameter that the function expects to receive. All the parameters for a function are input parameters and are nullable. For more information, see Defining the parameters in CREATE FUNCTION.

parameter-name

Names the parameter. The name is used to refer to the parameter within the body of the function. The name cannot be the same as any other parameter-name in the parameter list.

data-type1

Specifies the data type of the input parameter. The data type can be a built-in data type or a distinct data type.

built-in-type: Specifies a built-in data type. For a more complete description of each built-in data type, see CREATE TABLE.
distinct-type-name: Specifies a distinct type. The length, precision, or scale attributes for the parameter are those of the source type of the distinct type (those specified on CREATE TYPE). For more information about creating a distinct type, see CREATE TYPE (distinct).
If the name of the distinct type is unqualified, the database manager resolves the schema name by searching the schemas in the SQL path.

If a CCSID is specified, the parameter will be converted to that CCSID prior to passing it to the function. If a CCSID is not specified, the CCSID is determined by the default CCSID at the current server at the time the function is invoked.

default-clause

Specifies a default value for the parameter. The default can be a constant, a special register, a global variable, an expression, or the keyword NULL. The expression is any expression defined in Expressions, that does not include an aggregate function or column name. If a default value is not specified, the parameter has no default and cannot be omitted on invocation. The maximum length of the expression string is 64K.

The default expression must be assignment compatible to the parameter data type.

Any comma in the default expression that is intended as a separator of numeric constants in a list must be followed by a space.

All objects referenced in a default expression must exist when the function is created.

RETURNS TABLE

Specifies the output table of the function.

Start of change Assume the number of parameters is N. There must be no more than 8000-N columns. End of change

column-name

Specifies the name of a column of the output table. Do not specify the same name more than once.

data-type2

Specifies the data type and attributes of the output.

You can specify any built-in data type (except LONG VARCHAR, or LONG VARGRAPHIC) or a distinct type. When the function is invoked the results are assigned to these data types (using storage assignment rules).

If a CCSID is specified and the CCSID of the return data is encoded in a different CCSID, the data is converted to the specified CCSID.

If a CCSID is not specified and the function is not referenced in a view, the return data is converted to the CCSID of the job (or associated graphic CCSID of the job for graphic string return values), if the CCSID of the return data is encoded in a different CCSID. To avoid any potential loss of characters during the conversion, consider explicitly specifying a CCSID that can represent any characters that will be returned from the function. This is especially important if the data type is graphic string data. In this case, consider using CCSID 1200 or 13488 (Unicode graphic string data).

If a CCSID is not specified and the function is referenced in a view, the return data is converted to the CCSID of the associated view column. To avoid any potential loss of characters during the conversion, consider explicitly specifying a CCSID that can represent any characters that will be returned from the function. This is especially important if the data type is graphic string data. In this case, consider using CCSID 1200 or 13488 (Unicode graphic string data).

LANGUAGE SQL

Specifies that this is an SQL function.

SPECIFIC specific-name

Specifies a unique name for the function. For more information on specific names, see Specifying a specific name for a function in CREATE FUNCTION.

PROGRAM NAME external-program-name

Specifies the unqualified name of the service program to be created for the function. external-program-name must be a valid system name. End of change

GLOBAL DETERMINISTIC or STATEMENT DETERMINISTIC or NOT DETERMINISTIC

Specifies whether the function returns the same results each time that the function is invoked with the same input arguments. The default is NOT DETERMINISTIC.

NOT DETERMINISTIC: Specifies that the function might not return the same result each time that the function is invoked with the same input arguments. The function depends on some state values that affect the results. The database manager uses this information during optimization of SQL statements. An example of a table function that is not deterministic is one that references special registers, non-deterministic functions, or a sequence in a way that affects the table function result table.
GLOBAL DETERMINISTIC: Specifies that the function always returns the same result table each time that the function is invoked with the same input arguments. The database manager uses this information during optimization of SQL statements. The query optimizer may choose to cache global deterministic function results.
STATEMENT DETERMINISTIC: Specifies that the function might not return the same result each time that the function is invoked with the same input arguments, but multiple invocations of the function within a single SQL statement are considered deterministic. The query optimizer will not cache statement deterministic function results.¹

EXTERNAL ACTION or NO EXTERNAL ACTION

Specifies whether the function takes an action that changes the state of an object that the database manager does not manage. An example of an external action is sending a message or writing a record to a stream file. The default is EXTERNAL ACTION.

EXTERNAL ACTION: Specifies that the function can take an action that changes the state of an object that the database manager does not manage. Thus, the function must be invoked with each successive function invocation. EXTERNAL ACTION should be specified if the function contains a reference to another function that has an external action.
NO EXTERNAL ACTION: The function does not perform an external action. It need not be called with each successive function invocation.
NO EXTERNAL ACTION functions might perform better than EXTERNAL ACTION functions because they might not be invoked for each successive function invocation.

CONTAINS SQL, READS SQL DATA, or MODIFIES SQL DATA

Specifies the classification of SQL statements Start of change

and nested routines End of change

that the function can execute. The database manager verifies that the SQL statements issued by the function, Start of change

and all routines locally invoked by the function, End of change

are consistent with this specification. Start of change

The verification is not performed when nested remote routines are invoked. End of change

For the classification of each statement, see Characteristics of SQL statements. The default is READS SQL DATA. This option applies to any parameter default expressions.

READS SQL DATA: Specifies that the function can execute statements with a data access classification of READS SQL DATA, CONTAINS SQL, or NO SQL. The function cannot execute SQL statements that modify data.
CONTAINS SQL: Specifies that the function can execute only SQL statements with a data access classification of CONTAINS SQL or NO SQL. The function cannot execute any SQL statements that read or modify data.
MODIFIES SQL DATA: The function can execute any SQL statement except those statements that are not supported in any function.

RETURNS NULL ON NULL INPUT or CALLED ON NULL INPUT

Specifies whether the function is called if any of the input arguments is null at execution time.

RETURNS NULL ON NULL INPUT: Specifies that the function is not called if any of the input arguments is null. The result is an empty table, which is a table with no rows. RETURNS NULL ON NULL INPUT is the default.
CALLED ON NULL INPUT: Specifies that the function is to be invoked if any argument values are null. This specification means that the function must be coded to test for null argument values. The function can return an empty table, depending on its logic.

INHERIT SPECIAL REGISTERS

Specifies that existing values of special registers are inherited upon entry to the function.

STATIC DISPATCH

Specifies that the function is dispatched statically. All functions are statically dispatched.

FENCED or NOT FENCED

Specifies whether the SQL function runs in an environment that is isolated from the database manager environment. FENCED is the default.

FENCED

The function will run in a separate thread.

FENCED functions cannot keep SQL cursors open across individual calls to the function. However, the cursors in one thread are independent of the cursors in any other threads which reduces the possibility of cursor name conflicts.

NOT FENCED

The function may run in the same thread as the invoking SQL statement.

NOT FENCED functions can keep SQL cursors open across individual calls to the function. Since cursors can be kept open, the cursor position will also be preserved between calls to the function. However, cursor names may conflict since the UDF is now running in the same thread as the invoking SQL statement and other NOT FENCED UDFs.

NOT FENCED functions usually perform better than FENCED functions.

ALLOW PARALLEL or DISALLOW PARALLEL

Specifies whether the function can be run in parallel.

The default is DISALLOW PARALLEL if one or more of the following clauses are specified: NOT DETERMINISTIC, EXTERNAL ACTION, or MODIFIES SQL DATA, or if this is a pipelined table function. Otherwise, ALLOW PARALLEL is the default.

ALLOW PARALLEL: Specifies that the database manager can consider parallelism for the function. The database manager is not required to use parallelism on the SQL statement that invokes the function or on any SQL statement issued from within the function.
See the descriptions of NOT DETERMINISTIC, EXTERNAL ACTION, and MODIFIES SQL DATA for considerations that apply to specification of ALLOW PARALLEL.
DISALLOW PARALLEL: Specifies that the database manager must not use parallelism for the function.

CONCURRENT ACCESS RESOLUTION

Specifies whether the database manager should wait for data that is in the process of being updated. DEFAULT is the default.

DEFAULT: Specifies that the concurrent access resolution is not explicitly set for this function. The value that is in effect when the function is invoked will be used.
WAIT FOR OUTCOME: Specifies that the database manager is to wait for the commit or rollback of data in the process of being updated.
USE CURRENTLY COMMITTED: Specifies that the database manager is to use the currently committed version of the data when encountering data that is in the process of being updated.; When the lock contention is between a read transaction and a delete or update transaction, the clause is applicable to scans with isolation level CS (but not for CS KEEP LOCKS).

SYSTEM_TIME SENSITIVE

Determines whether references to system-period temporal tables in both static and dynamic SQL statements are affected by the value of the CURRENT TEMPORAL SYSTEM_TIME special register. YES is the default.

YES: References to system-period temporal tables are affected by the value of the CURRENT TEMPORAL SYSTEM_TIME special register.
NO: References to system-period temporal tables are not affected by the value of the CURRENT TEMPORAL SYSTEM_TIME special register.

NOT SECURED or SECURED

Specifies whether the function is considered secure for row access control and column access control.

NOT SECURED: Specifies that the function is considered not secure for row access control and column access control. This is the default.; When the function is invoked, the arguments of the function must not reference a column for which a column mask is enabled when the table is using active column access control.
SECURED: Specifies that the function is considered secure for row access control and column access control.; A function must be defined as secure when it is referenced in a row permission or a column mask.

WRAPPED obfuscated-statement-text

Specifies the encoded definition of the function. A CREATE FUNCTION statement can be encoded using the WRAP scalar function.

CARDINALITY bigint

This optional clause provides an estimate of the expected number of rows to be returned by the function for optimization purposes. Valid values for integer range from 0 to 9 223 372 036 854 775 807 inclusive.

If the CARDINALITY clause is not specified for a table function, the database manager will assume a finite value as a default.

Start of change A table function that returns a row every time it is called and never returns the end-of-table condition has infinite cardinality. A query that invokes such a function and requires an eventual end-of-table condition before it can return any data will not return unless interrupted. End of change

SET OPTION-statement

Specifies the options that will be used to create the function. These options also apply to any default value expressions. For example, to create a debuggable function, the following statement could be included:

SET OPTION DBGVIEW = *SOURCE

The default values for the options depend on the options in effect at create time. For more information, see SET OPTION.

The options CNULRQD, CNULIGN, COMPILEOPT, NAMING, and SQLCA are not allowed in the CREATE FUNCTION statement. CLOSQLCSR(*ENDACTGRP) is always used for SQL table functions. The following options are used when processing default value expressions: ALWCPYDTA, CONACC, DATFMT, DATSEP, DECFLTRND, DECMPT, DECRESULT, DFTRDBCOL, LANGID, SQLCURRULE, SQLPATH, SRTSEQ, TGTRLS, TIMFMT, and TIMSEP.

SQL-routine-body

Specifies a single SQL statement, including a compound statement. See SQL control statements for more information about defining SQL functions.

Start of change A non-pipelined table function must contain exactly one RETURN statement. A pipelined table function must contain at least one RETURN statement. The RETURN statement must be executed when the function is invoked. End of change

A call to a procedure that issues a CONNECT, SET CONNECTION, RELEASE, DISCONNECT, COMMIT, ROLLBACK and SET TRANSACTION statement is not allowed in a function.

ALTER PROCEDURE (SQL), ALTER FUNCTION (SQL scalar), and ALTER FUNCTION (SQL table) with a REPLACE keyword are not allowed in an SQL-routine-body.

Notes

General considerations for defining user-defined functions: See SQL control statements for general information about defining user-defined functions.

Function ownership: If SQL names were specified:

If a user profile with the same name as the schema into which the function is created exists, the owner of the function is that user profile.
Otherwise, the owner of the function is the user profile or group user profile of the thread executing the statement.

If system names were specified, the owner of the function is the user profile or group user profile of the thread executing the statement.

Function authority: If SQL names are used, functions are created with the system authority of *EXCLUDE on *PUBLIC. If system names are used, functions are created with the authority to *PUBLIC as determined by the create authority (CRTAUT) parameter of the schema.

If the owner of the function is a member of a group profile (GRPPRF keyword) and group authority is specified (GRPAUT keyword), that group profile will also have authority to the function.

Pipelined and non-pipelined functions: There are two types of SQL table functions. A table function which does not contain any PIPE statements within the SQL-routine-body is a non-pipelined table function. It contains one RETURN statement which returns a table. A pipelined table function is a table function which contains one or more RETURN statements with no return values and zero or more PIPE statements within the SQL-routine-body. It returns a table a row at a time. The two types of table functions are invoked in exactly the same way.

A PIPE statement returns a result row from the table function. To get the next row, control returns in the SQL-routine-body to the statement following the PIPE statement.

REPLACE rules: When a function is recreated by REPLACE:

Any existing comment or label is discarded.
Authorized users are maintained. The object owner could change.
Current journal auditing is preserved.

If the function is replaced and the function signature or result data types are altered, the results from any function, materialized query table, procedure, trigger, or view that references the function may be unpredictable. Any referenced objects should be recreated.

Creating the function: When an SQL function is created, the database manager creates a temporary source file that will contain C source code with embedded SQL statements. A *SRVPGM object is then created using the CRTSRVPGM command. The SQL options used to create the service program are the options that are in effect at the time the CREATE FUNCTION statement is executed. The service program is created with ACTGRP(*CALLER).

When an SQL function is created, the function's attributes are stored in the created service program object. If the *SRVPGM object is saved and then restored to this or another system, the attributes are used to update the catalogs.

Start of change If the PROGRAM NAME clause is provided, its name is used for the creation of the service program object. End of change Otherwise, the specific name is used to determine the name of the source file member and *SRVPGM object. If the specific name is a valid system name, it will used as the name of member and program. If the member already exists, it will be overlaid. If a program already exists in the specified library, a unique name is generated using the rules for generating system table names. If the specific name is not a valid system name, a unique name is generated using the rules for generating system table names.

Invoking the function: When an SQL function is invoked, it runs in the activation group of the calling program.

Inline functions: When an SQL table function is inlined, instead of invoking the function as part of a query, the fullselect in the RETURN statement of the function may be copied (inlined) into the query itself. Such a function is called an inline function. A table function is an inline function if:

The SQL function is defined as NO EXTERNAL ACTION.
The SQL-routine-body contains only a RETURN statement.
No column in the result table is the XML data type.
All objects referenced in the function exist when the function is created.
The SQL-routine-body does not contain a common table expression that references an input parameter.

An inline function is only copied (inlined) into a query if:

The query is eligible for the SQL Query Engine (SQE).
The function does not reference a table on a different server.
The function references an object and the authority attributes of the function and the query are compatible based on one of the following conditions:
- The function is defined to run under the user's authority (*USER).
- The query is running under the owner's authority (*OWNER) and the owner of the query is the same as the owner of the function.
- The query is running under the user's authority (*USER), and the user or the user's group profile is the same as the owner of the function.
Note: If the function is defined as FENCED, the query must not use adopted authority. If the query runs under the owner's authority (*OWNER) and the function runs under the user’s authority (*USER), the owner of the query must be the same as the user or the user's group profile.

When a function is inlined, some of the options specified when the function was created are ignored:

PARALLEL or NOT PARALLEL
MODIFIES SQL DATA
Commitment control level
CONCURRENT ACCESS RESOLUTION
ALWCPYDTA
ATOMIC or NOT ATOMIC

If a function is inlined and it contains a reference to a special register, the value of the special register will be the same as other references to the same special register in the query.

Obfuscated statements: A CREATE FUNCTION statement can be executed in obfuscated form. In an obfuscated statement, only the function name and parameters are readable followed by the WRAPPED keyword. The rest of the statement is encoded in such a way that it is not readable but can be decoded by a database server that supports obfuscated statements. Obfuscated statements can be produced by invoking the WRAP scalar function. Any debug options that are specified when the function is created from an obfuscated statement are ignored. A function that is created from an obfuscated statement cannot be restored to a release where obfuscation is not supported.

Dependent objects: An SQL routine is dependent on objects that are referenced in the SQL-routine-body. The names of the dependent objects are stored in catalog view SYSROUTINEDEP. If the object reference in the SQL-routine-body is a fully qualified name or, in SQL naming, if an unqualified name is qualified by the current schema, then the schema name of the object in SYSROUTINEDEP will be set to the specified name or the value of the current schema. Otherwise, the schema name is not set to a specific schema name. Start of change Unqualified function names, variable names, and type names will have a schema name of CURRENT PATH. End of change If the name is not set to an actual schema name, then DROP and ALTER statements will not be able to determine whether the routine is dependent on the object being altered or dropped.

Syntax alternatives: The following keywords are synonyms supported for compatibility to prior releases. These keywords are non-standard and should not be used:

The keywords VARIANT and NOT VARIANT can be used as synonyms for NOT DETERMINISTIC and DETERMINISTIC.
The keywords NULL CALL and NOT NULL CALL can be used as synonyms for CALLED ON NULL INPUT and RETURNS NULL ON NULL INPUT.
The keywords IS DETERMINISTIC may be used as a synonym for DETERMINISTIC.

Example

Define a table function that returns the employees in a specified department number.

CREATE FUNCTION DEPTEMPLOYEES (DEPTNO CHAR(3))
  RETURNS TABLE (EMPNO CHAR(6),
                 LASTNAME VARCHAR(15),
                 FIRSTNAME VARCHAR(12))
  LANGUAGE SQL
  READS SQL DATA
  NO EXTERNAL ACTION
  DETERMINISTIC
  DISALLOW PARALLEL
  RETURN
      SELECT EMPNO,LASTNAME,FIRSTNME
        FROM EMPLOYEE
        WHERE EMPLOYEE.WORKDEPT =DEPTEMPLOYEES.DEPTNO

¹ The DETERMINISTIC_UDF_SCOPE QAQQINI option can be used to get this same behavior for a GLOBAL DETERMINISTIC function. For more information, see the Database Performance and Query Optimization topic collection.