Thread: FUNCTION, TRIGGER and best practices

FUNCTION, TRIGGER and best practices

From
"Keith Worthington"
Date:
Hi All,

I have a function/trigger pair operating that move data from one table to
another after that data is inserted into the first table.

Can triggers and functions have the same name?
Is this a good practice?
Where should the function and trigger be stored?
In the same schema as the table they are connected to?

I will now be attempting to expand this function to move data from two source
tables to four or more target tables.  (Different source columns go to
different tables.)  Is it possible to do a transaction/commit/rollback inside
a function that is driven by a trigger so that if any portion of the transfer
fails it is all rolled back?

Thanks to all on this list that have helped me get this far. :-)

Kind Regards,
Keith

______________________________________________
99main Internet Services http://www.99main.com


Re: FUNCTION, TRIGGER and best practices

From
Josh Berkus
Date:
Keith,

> Can triggers and functions have the same name?

Yes.

> Is this a good practice?

It can be.   If the function does nothing but power the trigger, sure.   Or
you can use prefixes or suffixes to distinguish them.  For example, if you
had an audit trigger on the companies table, you might call the function
audit_companies_tf, the trigger audit_companies_tg.

> Where should the function and trigger be stored?

Um, in the database?

> In the same schema as the table they are connected to?

I'm not sure you CAN put the trigger in a different schema from the table.
It's a dependant object.

> I will now be attempting to expand this function to move data from two
> source tables to four or more target tables.  (Different source columns go
> to different tables.)  Is it possible to do a transaction/commit/rollback
> inside a function that is driven by a trigger so that if any portion of the
> transfer fails it is all rolled back?

Well, if any portion of a trigger fails, everything is rolled back; so is the
insert, for a before trigger.  So that might be the way to handle the
double-trigger situation .... except they'll be in two different tables, yes?
So that'll require an explicit transaction wrapping the two inserts.

--
Josh Berkus
Aglio Database Solutions
San Francisco

Re: FUNCTION, TRIGGER and best practices

From
Michael Fuhr
Date:
On Thu, Dec 16, 2004 at 11:05:20PM -0500, Keith Worthington wrote:

> Can triggers and functions have the same name?

The "Trigger Procedures" section of the "PL/pgSQL - SQL Procedural
Language" chapter in the documentation shows a trigger having the
same name as the function it calls.  And you could always try it.

> Is this a good practice?

I'd suggest using whatever names seem the most self-documenting for
a particular application.  Maybe somebody else will present arguments
advocating one style or another; I don't have strong opinions either
way.

Multiple triggers on a table will fire alphabetically by trigger
name, so firing order might affect how you name the triggers.

> Where should the function and trigger be stored?
> In the same schema as the table they are connected to?

If objects (functions, types, etc.) are dedicated to tables in a
particular schema then I usually create them in that schema to keep
everything together.  That can be handy for doing backups with
"pg_dump --schema schemaname".

> I will now be attempting to expand this function to move data from two source
> tables to four or more target tables.  (Different source columns go to
> different tables.)  Is it possible to do a transaction/commit/rollback inside
> a function that is driven by a trigger so that if any portion of the transfer
> fails it is all rolled back?

See the last paragraph of the "Structure of PL/pgSQL" section in
the "PL/pgSQL - SQL Procedural Language" chapter of the documentation,
as well as the "Errors and Messages" section in the same chapter.
Functions can't execute COMMIT or ROLLBACK, but they can raise an
exception to abort the current transaction.  In PostgreSQL 8.0
functions can trap errors, effectively allowing you to have a
subtransaction inside your function.  See "Trapping Errors" in the
"Control Structures" documentation for details.

--
Michael Fuhr
http://www.fuhr.org/~mfuhr/