Monday, September 13, 2010

A Better Method

Just posted a new Tip over on that deals with quickly creating a package that can manage all DML transactions against a table or tables.

I came across this feature in the days of APEX 4.0 EA1.  I immediately thought that it was a new APEX 4.0 feature - until I checked 3.2, and it was there too.  Basically, this allows you to easily create a table API which can be used to replace the built-in APEX DML processes.

Why would you do this?

If you wanted to create a more secure architecture, you may want to have a tableless schema - thus, the built in processes will do you little good.  This API can live in a separate schema, and facilitate all DML calls used to manage your data.

Since it's in a PL/SQL package, you can easily add additional security or data integrity checks before performing any of the DML transactions - another major reason to consider this approach.


Patrick Barel said...

Maybe you could check out the QDA (Quest Development Architecture) available in QCGU (Quest CodeGen Utility) at

Anonymous said...

Also someone did a long time ago but never...

Anonymous said...

also forgot his source code website too..

Jason D. Aughenbaugh said...

Thanks for the tip Scott I have had questions from people on this type of setup for some time, and no really good answers. Now this one bears a little experimentation. :")

Scott said...

I do remember now that this feature was built WITH Dan's TapiGen APIs. Dan - if you're watching - can you confirm?

- Scott -

Joel R. Kallman said...

>> I do remember now that this feature was built WITH Dan's TapiGen APIs. Dan - if you're watching - can you confirm?

Really? That's odd. The date on Dan's code is 2008, the APEX TABLE API generator was written in February 2003. And the code between them is completely dissimilar. Dan's is way more sophisticated.

Sorry, but I don't think that statement is accurate at all.

Scott said...

I stand corrected.

Thanks, Joel.

- Scott -

Ino said...


I saw your ODTUG webcast about this subject, and tried it out myself. I think that this API is really too simple to use. E.g. for an update it doesn't take into account the columns you actually want to change, so the others may turn up as NULL after the update.

Also, the API cannot (easily) be used in a tabular form.

One of the suggestions in the webinar was to use views with instead of triggers. That could be a better solution.


Scott said...


If you pass all parameters back to the API, then they should all be updated. If you NULL some of all of them out, well, then your data will also reflect that.

Keep in mind that the example that I used is just one example; you could likely build an API generator that is more robust and suits your specific needs in no time at all.

I am not a huge fan of Instead of Triggers, as there is no way to rollback anything, should something fail. For example: how would you handle lost update detection?


- Scott -