Usage of mainframe DB2 trigger

General Trigger Creation Syntax

$ cat create_trigger.sql

--#SET TERMINATOR #
DROP TRIGGER   #

CREATE TRIGGER 
  AFTER [UPDATE|INSERT|DELETE] ON 
  REFERENCING [NEW|NEW_TABLE] AS N
              [OLD|OLD_TABLE] AS O
  FOR EACH [ROW|STATEMENT] MODE DB2SQL
BEGIN ATOMIC
     {triggered-SQL-statement}
END#

$ db2 -f create_trigger.sql

Details of syntax please refer to IBM document

About Trigger Event Clause: [INSERT|UPDATE|DELETE]

Only one type could be provided, so if you want to listen on all modification types, you must create 3 triggers, one for INSERT, one for UPDATE, and one for DELETE.

About Transition Variables [NEW|NEW_TABLE|OLD|OLD_TABLE]

NEW|NEW_TABLE could only be used with INSERT and UPDATE trigger-events.

OLD|OLD_TABLE could only be used with UPDATE and DELETE trigger-events.

NEW and OLD could only be used with FOR EACH ROW.

See following tables for details:

Granularity	Activation Time	Triggering SQL Operation	Transition Variables Allowe	Transition Tables Allowed
FOR EACH ROW	BEFORE	INSERT	NEW	-
		UPDATE	OLD, NEW	-
		DELETE	OLD	-
	AFTER	INSERT	NEW	NEW TABLE
		UPDATE	OLD, NEW	OLD TABLE, NEW TABLE
		DELETE	OLD	OLD TABLE
OR EACH STATEMENT	BEFORE	INSERT	-	-
		UPDATE	-	-
		DELETE	-	-
	AFTER	INSERT	-	NEW TABLE
		UPDATE	-	OLD TABLE, NEW TABLE
		DELETE	-	OLD TABLE

NEW/OLD with "FOR EACH ROW"

• NEW means the new values of current INSERT or UPDATE row.

• OLD means the exact old value of current UPDATE or DELETE row.
  

NEW_TABLE/OLD_TABLE with "FOR EACH STATEMENT"

• NEW_TABLE means all the new changed rows of current INSERT or UPDATE statement affected.

• OLD_TABLE means all the new changed rows of current UPDATE or DELETE statement affected.
  

OLD_TABLE/NEW_TABLE with "FOR EACH ROW"

Usually when "FOR EACH ROW", you should refer to NEW and OLD, not NEW_TABLE, and OLD_TABLE; this is reasonable cases; but since DB2 provides the transition variable NEW_TABLE and OLD_TABLE with "FOR EACH ROW", you can also use them of course; however, they must be used carefully, because where OLD_TABLE and NEW_TABLE means all the changed rows, not the current NEW and OLD; for example: if you create a trigger as:

CREATE TRIGGER TESTTRIGGER
  AFTER UPDATE ON TESTTAB
  REFERENCING NEW_TABLE AS N
              OLD_TABLE AS O

  FOR EACH ROW MODE DB2SQL
BEGIN ATOMIC
              {refer to N and O}
END

And if an UPDATE statement update 10 rows data, this trigger will be fired 10 times, each time the transition variable NEW_TABLE and OLD_TABLE will contain 10 data rows.

About {triggered-SQL-statement}

A trigger body could include only SQL statements and built-in functions. So in a trigger body:
1.      You cannot declare local variable
2.      You cannot execute control statement, like IF, WHILE, etc
So if you want the trigger to perform actions or use logic that is not available in SQL statements or built-in functions, you need to write a stored procedure or a user defined function, and invoke it from the trigger body.

 This is very bad user experience.

To invoke a user defined function using VALUES statement, like
VALUES YOUR_FUNCTION(N.EMPNO,O.SALARY,N.SALARY);

To invoke a stored procedure using CALL statement, like
CALL YOUR_PROCEDURE (N.EMPNO,O.SALARY,N.SALARY);