Re: Alter project: client or server side?

> -----Original Message-----
> From: Jean-Michel POURE [mailto:jm(dot)poure(at)freesurf(dot)fr]
> Sent: 24 September 2001 20:11
> To: pgadmin-hackers(at)postgresql(dot)org
> Subject: [pgadmin-hackers] Alter project: client or server side?

< snipped long discussion about object dependencies and gotchas with regard
to dropping/recreating (== editting) of objects such as functions - e.g.
view frog uses function tadpole. Editting tadpole will break frog so we need
to rebuild frog as well >

> The way we do it in pgAdmin I
> http://cvs.social-housing.org/viewcvs.cgi/pgadmin1
> is that we maintain a dependency table based on STRING NAMES
> and not OIDs. When altering an object (view, function,
> trigger) we rebuild all dependent
> objects.
>
> Is this the way we should proceed with pgAdmin II?
> Is anyone planning a real dependency table based on object
> STRING NAMES?
>
> We need some advice:
> 1) Client solution: should we add the rebuilding feature to
> pgAdmin II?
> 2) Server solution: should we wait until the ALTER OBJECT
> project is complete?

I've CC'd this to pgsql-hackers in hope of some guidence from the developers
there.

My current view is that we need to implement these facilities (object
dependency tracking/rebuilding) client side. I believe we are just coming up
to the 7.2 beta and the required features do not exist to my knowledge,
therefore we either wait and hope they get written for 7.3 (or 8.0) or do it
ourselves client side.

Regards, Dave.