CREATE FUNCTION statement

The CREATE FUNCTION statement defines a callable function or procedure.

You can also use the CREATE PROCEDURE statement to define a callable function or procedure, also known as a routine.

SYNTAX

Notes:
  1. If the routine type is FUNCTION, the direction indicator (IN, OUT, or INOUT) is optional for each parameter. However, for documentation purposes, it is good programming practice to specify a direction indicator for all new routines; if you do not specify the direction, a default value of IN is used.
  2. When the NAMESPACE or NAME clause is used, its value is implicitly CONSTANT and of type CHARACTER. For information about the use of CONSTANT variables, see the DECLARE statement.
  3. If the routine type is FUNCTION, you cannot specify a LANGUAGE of DATABASE.

Overview

The CREATE FUNCTION and CREATE PROCEDURE statements define a callable function or procedure, also known as a routine.

In previous versions of this product, CREATE FUNCTION and CREATE PROCEDURE had different uses and different capabilities. Subsequent enhancements have resulted in the differences listed previously in notes 1 and 3.

Routines are useful for creating reusable blocks of code that can be run independently many times. You can implement them as a series of ESQL statements, a Java™ method, or a database stored procedure. This flexibility means that some of the clauses in the syntax diagram are not applicable (or allowed) for all types of routine.

Each routine has a name, which must be unique within the schema to which it belongs. Routine names therefore cannot be overloaded; if the broker detects that a routine name has been overloaded, it raises an exception.

The LANGUAGE clause specifies the language in which the routine's body is written. The options are:
DATABASE
The procedure is called as a database stored procedure.
ESQL
The procedure is called as an ESQL routine.
JAVA
The procedure is called as a static method in a Java class.
Unspecified
If you do not specify the LANGUAGE clause, the default language is ESQL unless you specify the EXTERNAL NAME clause (in which case, the default language is DATABASE).
Restrictions on the use of the LANGUAGE clause exist. You cannot use:
  • The ESQL option with an EXTERNAL NAME clause
  • The DATABASE or JAVA options without an EXTERNAL NAME clause
  • The DATABASE option with a routine type of FUNCTION

Specify the routine's name using the RoutineName clause, and the routine's parameters using the ParameterList clause. If the LANGUAGE clause specifies ESQL, implement the routine using a single ESQL statement. This statement is most useful if it is a compound statement (BEGIN ... END), because it can then contain as many ESQL statements as necessary to fulfil its function.

Alternatively, instead of providing an ESQL body for the routine, you can specify a LANGUAGE clause other than ESQL. You can then use the EXTERNAL NAME clause to provide a reference to the actual body of the routine, wherever it is located externally to the broker. For more information about using the EXTERNAL NAME clause, see Invoking stored procedures and Calling a Java routine.

Routines of any LANGUAGE type can have IN, OUT, and INOUT parameters. The caller can pass several values into the routine, and receive back several updated values. These returned parameters are in addition to any RETURNS clause that you have defined for the routine. The RETURNS clause defines the value that the routine returns to the caller.

Routines that are implemented in different languages have their own restrictions on which data types can be passed in or returned; these restrictions are documented later in this section. The data type of the returned value must match the data type of the value that is defined to be returned from the routine. Also, if a routine is defined to have a return value, the caller of the routine cannot ignore it. For more information see CALL statement.

Routines can be defined in either a module or a schema. Routines that are defined in a module are local in scope to the current node, which means that only code belonging to that same module (or node) can invoke them. Routines that are defined in a schema, however, can be invoked by using either of the following options:
  • Code in the same schema
  • Code in any other schema, if either of the following conditions applies:
    • The other schema's PATH clause contains the path to the called routine
    • The called routine is invoked using its fully qualified name (which is its name, prefixed by its schema name, separated by a period)
Thus, if you need to invoke the same routine in more than one node, define it in a schema.

For any language or routine type, the method of invocation of the routine must match the manner of declaration of the routine. If the routine has a RETURNS clause, use either the FUNCTION invocation syntax or a CALL statement with an INTO clause. Conversely, if a routine has no RETURNS clause, you must use a CALL statement without an INTO clause.

Parameter directions

Parameters that are passed to routines always have a direction associated with them, which is one of the following types:
IN
The value of the parameter cannot be changed by the routine. A NULL value for the parameter is allowed, and can be passed to the routine.
OUT
When it is received by the called routine, the parameter that is passed into the routine always has a NULL value of the correct data type. This value is set irrespective of its value before the routine is called. The routine is allowed to change the value of the parameter.
INOUT
INOUT is both an IN and an OUT parameter. It passes a value into the routine, and the value that is passed in can be changed by the routine. A NULL value for the parameter is allowed, and can be passed both into and out of the routine.

If the routine type is FUNCTION, the direction indicator (IN, OUT, INOUT) is optional for each parameter. However, it is good programming practice to specify a direction indicator for all new routines of any type for documentation purposes.

ESQL variables that are declared to be CONSTANT (or references to variables declared to be CONSTANT) are not allowed to have the direction OUT or INOUT.

ESQL routines

ESQL routines are written in ESQL, and have a LANGUAGE clause of ESQL. The body of an ESQL routine is typically a compound statement of the form BEGIN … END, that contains multiple statements for processing the parameters that are passed to the routine.

ESQL example 1

The following example shows the same procedure as in Database routine example 1, but is implemented as an ESQL routine and not as a stored procedure. The CALL syntax and results of this routine are the same as those in Restrictions on Java routines.
CREATE PROCEDURE swapParms (
  IN parm1 CHARACTER,
  OUT parm2  CHARACTER,
  INOUT parm3 CHARACTER )
BEGIN
   SET parm2 = parm3;
   SET parm3 = parm1;
 END; 

ESQL example 2

This example procedure shows the recursive use of an ESQL routine. It parses a tree, visiting all places at and below the specified starting point, and reports what it has found:

SET OutputRoot.MQMD = InputRoot.MQMD;

  DECLARE answer CHARACTER;
  SET     answer = '';

  CALL navigate(InputRoot.XMLNS, answer);
  SET OutputRoot.XMLNS.Data.FieldNames = answer;


  CREATE PROCEDURE navigate (IN root REFERENCE, INOUT answer CHARACTER)
  BEGIN
    SET answer = answer || 'Reached Field... Type:' 
    || CAST(FIELDTYPE(root) AS CHAR)||
    ': Name:' || FIELDNAME(root) || ': Value :' || root || ': ';

    DECLARE cursor REFERENCE TO root;
    MOVE cursor FIRSTCHILD;
    IF LASTMOVE(cursor) THEN
      SET answer = answer || 'Field has children... drilling down ';
    ELSE
      SET answer = answer || 'Listing siblings... ';
    END IF;

    WHILE LASTMOVE(cursor) DO
      CALL navigate(cursor, answer);
      MOVE cursor NEXTSIBLING;
    END WHILE;

    SET answer = answer || 'Finished siblings... Popping up ';
  END;

When given the following input message:

<Person>
  <Name>John Smith</Name>
  <Salary period='monthly' taxable='yes'>-1200</Salary>
</Person>

the procedure produces the following output, which has been manually formatted:

  Reached Field... Type:16777232: Name:XML: Value :: Field has children... 
  drilling down 
  Reached Field... Type:16777216: Name:Person: Value :: Field has children...
  drilling down 
  Reached Field... Type:16777216: Name:Name: 
  Value :John Smith: Field has children... drilling down 
  Reached Field... Type:33554432: Name:: 
  Value :John Smith: Listing siblings... Finished siblings... Popping up
  Finished siblings... Popping up 
  Reached Field... Type:16777216: Name:Salary:
  Value :-1200: Field has children... drilling down 
  Reached Field... Type:50331648: Name:period: 
  Value :monthly: Listing siblings... Finished siblings... Popping up
  Reached Field... Type:50331648: Name:taxable: 
  Value :yes: Listing siblings... Finished siblings... Popping up 
  Reached Field... Type:33554432: Name:: 
  Value :-1200: Listing siblings... Finished siblings... Popping up 
  Finished siblings... Popping up 
  Finished siblings... Popping up 
  Finished siblings... Popping up

Java routines

A Java routine is implemented as a Java method, and has a LANGUAGE clause of JAVA. For Java routines, the ExternalRoutineName must contain the class name and method name of the Java method to be called. Specify the ExternalRoutineName like this:
>>--"-- className---.---methodName--"--------------><
where className identifies the class that contains the method and methodName identifies the method to invoke. If the class is part of a package, the class identifier part must include the complete package prefix; for example, "com.ibm.broker.test.MyClass.myMethod".

To find the Java class, the broker uses the search method that is described in Deploying Java classes.

Any Java method that you want to invoke must have the following basic signature:
public static <return-type> <method-name> (< 0 - N parameters>)

where <return-type> must be in the list of Java IN data types in the table in ESQL to Java data type mapping (excluding the REFERENCE type, which is not permitted as a return value), or the Java void data type. The parameter data types must also be in the ESQL to Java data type mapping table. In addition, the Java method is not allowed to have an exception throws clause in its signature.

The Java method’s signature must match the ESQL routine’s declaration of the method. You must also observe the following rules:
  • Ensure that the Java method name, including the class name and any package qualifiers, matches the procedure's EXTERNAL NAME.
  • If the Java return type is void, do not put a RETURNS clause on the ESQL routine’s definition. Conversely, if the Java return type is not void, you must put a RETURNS clause on the ESQL routine’s definition.
  • Ensure that every parameter's type and direction matches the ESQL declaration, according to the rules listed in the table in ESQL to Java data type mapping.
  • Ensure that the method's return type matches the data type of the RETURNS clause.
  • Enclose EXTERNAL NAME in quotation marks because it must contain at least "class.method".
  • If you want to invoke an overloaded Java method, you must create a separate ESQL definition for each overloaded method and give each ESQL definition a unique routine name.

You can use the Java user-defined node API in your Java method, provided that you observe the restrictions documented in Restrictions on Java routines. For more information about using the Java API, see Compiling a Java user-defined node.

Java routine example 1

This routine contains three parameters of varying directions, and returns an integer, which maps to a Java return type of java.lang.Long.

CREATE FUNCTION  myProc1( IN P1 INTEGER, OUT P2 INTEGER, INOUT P3 INTEGER )
 RETURNS INTEGER
 LANGUAGE JAVA 
 EXTERNAL NAME "com.ibm.broker.test.MyClass.myMethod1";

You can use the following ESQL to invoke myProc1:

CALL myProc1( intVar1, intVar2, intVar3) INTO intReturnVar3;
-- or
SET intReturnVar3 = myProc1( intVar1, intVar2, intVar3);

Java routine example 2

This routine contains three parameters of varying directions and has a Java return type of void.

CREATE PROCEDURE myProc2( IN P1 INTEGER, OUT P2 INTEGER, INOUT P3 INTEGER )
 LANGUAGE JAVA 
 EXTERNAL NAME "com.ibm.broker.test.MyClass.myMethod2";

You must use the following ESQL to invoke myProc2:

CALL myProc2(intVar1, intVar2, intVar3);

The following Java class provides a method for each of the preceding Java examples:

package com.ibm.broker.test;

class MyClass {
public static Long myMethod1( Long P1, Long[] P2 Long[] P3) { ... }
public static void myMethod2( Long P2, Long[] P2 Long[] P3) { ... }

 /* When either of these methods is called:
    P1 may or may not be NULL (depending on the value of intVar1).
    P2[0] is always NULL (whatever the value of intVar2).
    P3[0] may or may not be NULL (depending on the value of intVar3).  
    This is the same as with LANGUAGE ESQL routines. 
    When these methods return:
         intVar1 is unchanged
         intVar2 may still be NULL or may have been changed
         intVar3 may contain the same value or may have been changed.
     This is the same as with LANGUAGE ESQL routines.
     
    When myMethod1 returns: intReturnVar3 is either  NULL (if the
    method returns NULL) or it contains the value returned by the 
    method.
 */
}

ESQL to Java data type mapping

The following table summarizes the mappings from ESQL to Java.
Notes:
  • Only the Java scalar wrappers are passed to Java.
  • The ESQL scalar types are mapped to Java data types as object wrappers, or object wrapper arrays, depending upon the direction of the procedure parameter. Each wrapper array contains exactly one element.
  • Scalar object wrappers are used to allow NULL values to be passed to and from Java methods.
ESQL data types 1 Java IN data types Java INOUT and OUT data types
INTEGER, INT java.lang.Long java.lang.Long []
FLOAT java.lang.Double java.lang.Double[]
DECIMAL java.math.BigDecimal java.math.BigDecimal[]
CHARACTER, CHAR java.lang.String java.lang.String[]
BLOB byte[] byte[][]
BIT java.util.BitSet java.util.BitSet[]
DATE com.ibm.broker.plugin.MbDate com.ibm.broker.plugin.MbDate[]
TIME 2 com.ibm.broker.plugin.MbTime com.ibm.broker.plugin.MbTime[]
GMTTIME 2 com.ibm.broker.plugin.MbTime com.ibm.broker.plugin.MbTime[]
TIMESTAMP 2 com.ibm.broker.plugin.MbTimestamp com.ibm.broker.plugin.MbTimestamp[]
GMTTIMESTAMP 2 com.ibm.broker.plugin.MbTimestamp com.ibm.broker.plugin.MbTimestamp[]
INTERVAL Not supported Not supported
BOOLEAN java.lang.Boolean java.lang.Boolean[]
REFERENCE (to a message tree) 3 4 5 6 com.ibm.broker.plugin.MbElement com.ibm.broker.plugin.MbElement[] (Supported for INOUT. Not supported for OUT)
ROW Not supported Not supported
LIST Not supported Not supported
  1. Variables that are declared to be CONSTANT (or references to variables that are declared to be CONSTANT) are not allowed to have the direction INOUT or OUT.
  2. The time zone set in the Java variable is not important; you obtain the required time zone in the output ESQL.
  3. The reference parameter cannot be NULL when passed into a Java method.
  4. The reference cannot have the direction OUT when passed into a Java method.
  5. If an MbElement is passed back from Java to ESQL as an INOUT parameter, it must point to a location in the same message tree as that pointed to by the MbElement that was passed into the called Java method.

    For example, if an ESQL reference to OutputRoot.XML.Test is passed into a Java method as an INOUT MbElement, but a different MbElement is passed back to ESQL when the call returns, the different element must also point to somewhere in the OutputRoot tree.

  6. An MbElement cannot be returned from a Java method with the RETURNS clause, because no ESQL routine can return a reference. However, an MbElement can be returned as an INOUT direction parameter, subject to the conditions described in point 5.

A REFERENCE to a scalar variable can be used in the CALL of a Java method, provided that the data type of the variable to which the reference refers matches the corresponding data type in the Java program signature.

Restrictions on Java routines

The following restrictions apply to Java routines that are called from ESQL:
  • The Java method must be threadsafe (reentrant).
  • Database connections must be JDBC type 2 or type 4. Furthermore, database operations are not part of a broker transaction and therefore cannot be controlled by an external resource coordinator (as is the case in an XA environment).
  • The Java user-defined node API must be used only by the same thread that invoked the Java method.

    You can create threads inside your method. However, created threads must not use the Java APIs, and you must return control back to the broker.

    All restrictions that apply to the usage of the Java API also apply to Java methods that are called from ESQL.

  • Java methods that are called from ESQL must not use the MbNode class. Therefore, they cannot create objects of type MbNode, or call any of the methods on an existing MbNode object.
  • WebSphere® MQ or JMS work done inside a Java method that is called from ESQL must be done in accordance with the guidelines for performing WebSphere MQ and JMS work in a user-defined node. Start of changeSee Planning user-defined input nodes.End of change

Deploying Java classes

You can deploy your Java classes to a broker within a Java Archive (JAR) file, using one of the following two methods:
  1. Add the JAR file to the broker archive (BAR) file

    The most efficient and flexible method of deploying to the broker is to add your JAR file to the BAR file. You can do this manually or automatically using the workbench.

    If the workbench finds the correct Java class inside a referenced Java project open in the workspace, it automatically compiles the Java class into a JAR file and adds it to the BAR file. This procedure is the same procedure that you follow to deploy a JavaCompute node inside a JAR, as described in User-defined node classloading.

    When you deploy a JAR file from the workbench, the flow that has been redeployed reloads the JAR file contained in the BAR file.

    The files are also reloaded if the message flow that references a Java class is stopped and restarted. Ensure that you stop and restart (or redeploy) all flows that reference the JAR file that you want to update. This action avoids the problem of some flows running with the old version of the JAR file and other flows running with the new version.

    The workbench deploys only JAR files; it does not deploy standalone Java class files.

  2. Store the JAR file in either of the following locations:
    1. The workpath/shared-classes/ folder on the machine running the broker
    2. The CLASSPATH environment variable on the computer running the broker

    You must complete this action manually; you cannot use the workbench.

    In this method, redeploying the message flow does not reload the referenced Java classes; neither does stopping and restarting the message flow. The only way to reload the classes in this case is to stop and restart the broker itself.

To enable the broker to find a Java class, ensure that it is in one of the preceding locations. If the broker cannot find the specified class, it generates an exception.

Although you have the choices shown previously when you deploy the JAR file, using the workbench to deploy the BAR file provides the greatest flexibility when redeploying the JAR file.

Database routines

CREATE FUNCTION does not support database routines. Use CREATE PROCEDURE to define a database routine.

Related concepts
ESQL overview
Related tasks
Developing ESQL
Invoking stored procedures
Compiling a Java user-defined node
Adding files to a broker archive
Related reference
Syntax diagrams: available types
ESQL statements
CALL statement
CREATE PROCEDURE statement
ESQL-to-Java data-type mapping table
Related information
Java user-defined extensions API
Notices | Trademarks | Downloads | Library | Support | Feedback

Copyright IBM Corporation 1999, 2009Copyright IBM Corporation 1999, 2009.
Last updated : 2009-01-07 15:22:16

ak04960_