Saturday, January 30, 2010

Look, But Don't Touch

Joel Kallman had an interesting post about a customer who broke all of the rules and modified some constraints in the APEX_030200 schema.  In this case, there was a happy ending, as Joel went the extra mile to craft a custom patch to remedy the issue.  However, things don't always end up this way.

This reminds me of my early days with Oracle, when I used to work with the eBusiness Suite.  The first page of any documentation of the suite said something like this: "WARNING: Do not use SQL*Plus to modify any database object directly".  We used to laugh at this, as we used SQL*Plus & TOAD to modify these database objects several times a day, if not more frequently.

However, this was done on internal systems used to build demos.  If we messed something up - and boy did we ever - we could easily refresh the instance with little effort.

Fast-forward a few years to today:  FND & APPLSYS have been replaced by a single, much smaller schema called APEX_030200.  Despite this, the same rules apply:  NEVER modify anything in the APEX_030200 schema.  Period.  I always make it a point to emphasize this as much as possible during our training classes.

Now, if you MUST take a look at that schema - and many of you reading this will - please do it in a test environment that is not used for anything else.  This can easily be done by using VMWare or one of the many other virtual machine programs.  You can also do this in the Amazon Cloud.  It doesn't matter as much WHERE you set up a test instance, but rather that you DO set up a test instance, and then poke around there.  This way, nothing gets corrupted - now or in the future.


Gary Myers said...

I recall amending Forms 3.0 inp files in an editor rather than through the application.

The point of rules is to make you think twice about what you are doing before you break them

Byte64 said...

I agree in principle, however...

I mean, there should not be any need to stress the fact that who is changing things in the database using unsupported methods is embarking on an adventure of his own and there is no obligation for Oracle to restore things to a prior state. It's like saying if you jump off a cliff, you'd better off bringing your own parachute, don't expect the landing to be soft.

With this clear statement in mind, i spent most of the last 25 years "poking around" because that is the only way to understand things more deeply, it's the black box approach.

As you remarked however, this must be done in a controlled environment, where you can easily revert the system to its working state.

Clearly sometimes i happened to forgot to take a backup first... :-D

Scott said...

I think we can all say that we've all modified something that we shouldn't have at some point in our careers. Sometimes it worked; others, well... not so much.

The main points that you both brought up is that

1) don't do it
2) when break rule #1, be REALLY careful and
3) understand that if you break it, you bought it, so to speak

As nice and good as the APEX team is, I don't think that they can afford to be in the business of writing custom patches to bail out customers time and time again.

- Scott -

carpenteri said...

Thanks John for the interesting post. (and for your half of the excellent book "Pro Application Express).

I've recently encountered an issue within Apex which at the moment I can't see a way around without doing what you have said not to do!

I posted the question over on stackoverflow (

Is this the exception to the rule? or is there a better way to do this?