Re: sql_drop Event Trigger

Robert Haas <robertmhaas(at)gmail(dot)com> writes:
> Well, there's this, upon which we surely have not achieved consensus:
>
> http://www.postgresql.org/message-id/CA+TgmobQ6NGsxGuiHWqcygF0Q+7Y9zHNERePo3S1vsWKKNw2TQ@mail.gmail.com

Sub-Transaction Handling. I fail to come up with a regression test
showing any problem here, and Álvaro is now working on the patch so he
might be finding both a failure test and a fix.

> And then Tom also wrote this, which is kind of a good point, too:
>
>> Well, a list of object OIDs is of exactly zero use once the command
>> has been carried out. So I don't think that that represents a useful
>> or even very testable feature on its own, if there's no provision to
>> fire user code while the OIDs are still in the catalogs.

That's the reason why I've been proposing that we first add some
information to the event triggers, then see about the DROP support.

You might want to realize that the current event triggers implementation
is not even publishing the object ID now, only the command tag and the
name of the event.

We still don't have any way to make any use of the whole thing, apart
from maybe "block all commands" (but all is in fact a subset of known
DDL). I don't think we still are able to solve *any* use case with
what's currently commited. At all.

Given that, it's really hard for me to get excited on that very point.

Regards,
--
Dimitri Fontaine
http://2ndQuadrant.fr PostgreSQL : Expertise, Formation et Support