Search Postgresql Archives

Re: Event trigger information accessibility on plpgsql

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

 



Thanks Andrew,

You think it will be possible to, instead of comparing schemas, looking for the last modified OID on the DB to figure out where it happened?



On Jul 29, 2013, at 3:42 PM, Andrew Tipton wrote:

On Mon, Jul 29, 2013 at 9:12 PM, Javier de la Torre <jatorre@xxxxxxxxxxxxxx> wrote:
Hi, 

I was looking at the new event triggers on 9.3 and was doing some testing. I have compiled PostgreSQL 9.3 Beta2
It looks from the examples that the only info right now available on plpgsql when the triggers fire is tg_event and tg_tag.

When developing on C you get access to more things. But would it be possible to do a plpgsql trigger where I have access to the table name and OID when I do an ALTER table to rename a column? Right now it feels i can only know that a table has been altered, but not which one

I recall that the event triggers functionality was quite large, and hence was split across a series of patches.  Only some of those patches were able to be committed in time for 9.3, hence the lack of information accessible from plpgsql functions.  [The crux of the issue preventing the remaining patches from landing was, I believe, disagreement over how to expose the additional information in a consistent manner.]

Here is a crazy idea that might work:  create an event trigger which is fired on any CREATE or ALTER TABLE command (for any table), and in the trigger function compare the current state of the catalog (pg_class and pg_attribute) with a "snapshot" of the previous catalog state.  If anything has changed, perform the appropriate actions and update the "snapshot" with the new state.

The downside is that this might be quite slow.  But how often do you plan to be creating and altering tables?  The full scans of the catalog tables will only happen when CREATE TABLE or ALTER TABLE commands are executed, which might be acceptable.  And when additional event trigger information is presumably added in 9.4, you can simply treat it as a performance optimization.

[I'm toying with the idea of an extension which (ab)uses event triggers in precisely this manner.  The goal is to provide built-in version control that doesn't need any external tools to be run after changes have been made to the schema.]


Regards,
Andrew Tipton


[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]
[Index of Archives]     [Postgresql Jobs]     [Postgresql Admin]     [Postgresql Performance]     [Linux Clusters]     [PHP Home]     [PHP on Windows]     [Kernel Newbies]     [PHP Classes]     [PHP Books]     [PHP Databases]     [Postgresql & PHP]     [Yosemite]
  Powered by Linux