Kris Jurka wrote:
> Andreas Pflug wrote:
>> It's _always_ fine, because GetSchema obeys the search path and the
>> reengineered SQL is meant to be used in a search path situation as it
>> was at the time of reengineering. There are plenty of other
>> situations where the reduced form (omitting search-pathed schema)
>> won't work.
> Let me back up and make clear what I'm saying. The code I originally
> submitted is not correct. The code you committed is not correct
> either. To correctly determine what the default value for a serial
> will look like, you need to know what the search_path was at table
> creation time. Knowing its current value is not relevent.
> Run the following in psql.
> SET search_path TO public;
> CREATE schema s1;
> CREATE TABLE s1.t1(a serial);
> SET search_path TO s1;
> CREATE TABLE s1.t2(a serial);
> \d t1
> \d t2
> Note how one default includes the schema and the other doesn't.
> Explain how pgadmin can correctly determine the default value for both
> of these tables.
The problem of non-standard serials using pg_depend is on the TODO-list.
AFAICS the current implementation is better than before, and I don't
think adding more brains to this inferiour pattern matching approach
wouldn't make things better.
In response to
pgadmin-hackers by date
|Next:||From: August Zajonc||Date: 2006-02-20 22:49:00|
|Subject: Small update to README.admin81|
|Previous:||From: Dave Page||Date: 2006-02-20 20:33:47|
|Subject: Re: Query tool results in grid|