Can anyone kindly explain what mutative triggers are in Oracle?

Question

hi
Can anyone kindly explain what mutative triggers are in Oracle with example?
what is frag in oracle?

Answer 1

Oracle raises the mutating table error to protect you from building in-deterministic software.
Let’s explain that with a very simple example. Here’s a simple EMP-table:
ENAME      SAL
======     ====
SMITH      6000
JONES      4000Let’s suppose you have a business rule that dictates that the average salary is not allowed to exceed 5000. Which is true for above EMP-table (avg. SAL is exactly 5000).
And you have (erroneously) built after-row triggers (insert and update) to verify this business rule. In your row trigger you compute the average salary, check it’s value and if it’s more than 5000 you raise an application error.
Now you issue following DML-statement, to increase the salary of employees earning less than the average salary, and to decrease the salary of employees earning more than the average salary.
Update EMP
Set SAL = SAL + ((select avg(SAL) from EMP) – SAL)/2;The end result of this update is:
ENAME      SAL
======     ====
SMITH      5500
JONES      4500Note that the business rule is still OK: the average salary still doesn’t exceed 5000. But what happens inside your row-trigger, that has a query on the EMP-table?
Let’s assume the rows are changed in the order I displayed them above. The first time your row trigger fires it sees this table:
ENAME      SAL
======     ====
SMITH      5500
JONES      4000The row trigger computes the average and sees that it is not more than 5000, so it does not raise the application error. The second time your row trigger fires it sees the end result.
ENAME      SAL
======     ====
SMITH      5500
JONES      4500For which we already concluded that the row-trigger will not raise the application error.
But what happens if Oracle in it’s infinite wisdom decides to process the rows in the other order? The first time your row trigger executes it sees this table:
ENAME      SAL
======     ====
SMITH      6000
JONES      4500And now the row-trigger concludes that the average salary exceeds 5000 and your code raises the application error.
Presto. You have just implemented indeterministic software. Sometimes it will work, and sometimes it will not work.
Why? Because you are seeing an intermediate snapshot of the EMP-table, that you really should not be seeing (that is: querying).
This is why Oracle prevents you from querying the table that is currently being mutated inside row-triggers (i.e. DML is executed against it).
It’s just to protect you against yourself.
(PS1. Note that this issue doesn't happen inside statement-triggers)
(PS2. This also shows that mutating table error is really only relevant when DML-statements affect more that one row.)
Edited by: Toon Koppelaars on Apr 26, 2010 11:29 AM