Обсуждение: plperl and triggers

The manual says that $_TD->{old} contains the old value of the row
being operated on in a trigger, and $_TD->{new} the new value to be
used istead.

When using plpgsql I find that the equivs OLD and NEW do work that
way, but with plperlu there was only $_TD->{new}.

Perhaps {old} is only set on an update.  Or on an AFTER trigger to
an insert?

Or perhaps {old} only gets set if {new} is altered?

Whatever the reality, the doc should be fixed to explain exactly
when {old} and {new} exist and their semantics.

Liberal use of, eg, elog(WARNING, Dumper(\$_TD)) made it easy to
get my trigger function working as desired, but it would be cool
to document the details....

(In general, the docs are great.  I do sometimes have to read a
given section more than once -- or thrice -- to get it, but that
is as much due to fatigue as anything.  And I am one of those who
tend to learn new stuff better when reading from  paper than from
emitive displays like lcds or crts, for whatever reason.)

(Or maybe I'm just getting old. ;-)

-JimC
--
James Cloos <cloos@jhcloos.com>         OpenPGP: 0xED7DAEA6

On Sun, Sep 10, 2006 at 11:41:38PM -0400, James Cloos wrote:
> The manual says that $_TD->{old} contains the old value of the row
> being operated on in a trigger, and $_TD->{new} the new value to be
> used istead.
>
> When using plpgsql I find that the equivs OLD and NEW do work that
> way, but with plperlu there was only $_TD->{new}.

Are you sure the tests were identical other than the trigger function
language?  Based on your description I'd guess that you did an UPDATE
with PL/pgSQL and an INSERT with PL/PerlU.

> Perhaps {old} is only set on an update.  Or on an AFTER trigger to
> an insert?
>
> Or perhaps {old} only gets set if {new} is altered?

New and/or old rows are available in row-level BEFORE and AFTER
triggers.  New rows are available in INSERT and UPDATE triggers;
old rows are available in UPDATE and DELETE triggers.  This behavior
should be the same in all languages; if you have an example that
shows otherwise then please post it.

> Whatever the reality, the doc should be fixed to explain exactly
> when {old} and {new} exist and their semantics.

"A patch would be welcome" :-)  The PL/Perl documentation doesn't
describe the semantics but the "Triggers" chapter does.  If you've
found a discrepancy between documentation and behavior then please
post a self-contained test case.

--
Michael Fuhr

>>>>> "Michael" == Michael Fuhr <mike@fuhr.org> writes:

Michael> Are you sure the tests were identical other than the trigger
Michael> function language?  Based on your description I'd guess that
Michael> you did an UPDATE with PL/pgSQL and an INSERT with PL/PerlU.

I was sure when I wrote that, but -- most embarrassingly -- I've not
been able to reproduce....

In the exhausted euphoria of finishing a project I clearly [SIGH]
misremembered the steps along the way.

Apologies for the false alarm.

-JimC