Tuesday, November 22, 2011

So what triggers are we talking about?

Here's an overview of the talk.

This post will deal with the first bullet: the scope, what triggers are we talking about? And what triggers are we not talking about. Then there will probably be a few posts on 'properties' of the triggers, most notably I will spend some time on explaining the infamous mutating table error. Next we move on to a high level classification of use-cases of triggers. And talk a bit about why some of these might be considered harmful. Finally we will explain, in detail, the one use case where triggers are the perfect means to achieve the end.

The most common types of triggers, the ones everybody probably used at some time in their pl/sql programming career, are the "DML event" triggers. As above slide shows, there are twelve of such triggers: four each for every type of DML statement, Insert, Update and Delete. These triggers will be fired by the DBMS before a DML statement, after a DML statement, and before/after each affected row of the triggering DML statement. Stuff you all know right? The big difference between the statement-level and the row-level triggers, is that the latter ones can inspect (and change) the column-values of the current row that is being inserted/deleted/updated.

So here's an example. Suppose we have an EMP table that holds employees, and we want to execute an update statement that will increase the salary of all clerks (see update statement above). This will for the given table affect three rows. The before update statement trigger will then fire once. Next for each affected row the before and after row triggers will fire. And finally the after statement trigger will fire.

So if we create the four update triggers on the EMP table as follows:

We will get the following output (given we have set serveroutput to on).

Nothing new so far, I hope. Before we continue I just wanted to mention that as of Oracle11G we have the compound trigger feature.

 A compound trigger enables us to create the four update triggers above all in one go as follows:

Now, do you know why Oracle introduced compound triggers? I'll talk about that in a later post. What I'll say now is this: compound triggers are the answer of an enhancement request made by you (the pl/sql community) a long time ago. Because you have always hit a certain programming pattern with regards to triggers, when using them for a certain use case. Again I'll explain this in more detail in a future post.

So these are the triggers that are in-scope of this blog: DML event triggers, be them created individually or four in one go using the compound trigger mechanism.

Oracle DBMS offers us with many more triggers:

All of which will not be the matter of subject for this blog.

Stay tuned.


  1. Nice to see you got time to start this blog. Your presentation is most certainly a reason to attend a conference.

    The avoiding mutating table issue is most certainly something an Oracle programmer shoud have education about. At least until Oracle introduces exclusion constraints that is available in PostgreSQL. Looking forward for your future posts.

  2. Hei Rafu,

    Yes I have read about Jeff Davis' exclusion constraints in Postgres. But still they only solve a class of constraints.
    What we really need is CREATE ASSERTION from the Ansi/Iso SQL standard...

    But I'm getting ahead of things now.


  3. I applaud the publication of your article on triggers. It's a good reminder to look on the DevOps training.

    It is recommended to take DevOps training in Chennai quora

    Thank you for sharing with us the post that you have worked so hard to refine.

    DevOps training in chennai with placement | Best DevOps training in chennai | DevOps training in chennai OMR | DevOps training in chennai Velachery |DevOps training in chennai

  4. It is a great post. Keep sharing such kind of useful information.

    Guest posting sites