attlognum - Mailing list pgsql-hackers

From Christopher Kings-Lynne
Subject attlognum
Date
Msg-id GNELIHDDFBOCMGBFGEFOMEEDCEAA.chriskl@familyhealth.com.au
Whole thread Raw
List pgsql-hackers
I've been thinking about the 'attlognum' idea of a logical column number in
pg_attribute.

Back when I think Tom said 'we should do it now or never', I didn't think
too hard about it because I didn't realise the problem it would solve with
changing column type, and also how easy it would be to do.

It occurs to me now that it would be absolutely trivial to implement, as
opposed to being lots of work as I previously thought.  For 7.3, all we
would have had to do is add the column and then just make it equal attnum at
column creation time.  No other action is required except to tell ppl who
make admin software to ORDER BY attlognum rather than attnum.

Then, when we actually implement changing column type (a very useful
feature) we could start working with attlognums then and make sure '*'
expands in attlognum order rather than attnum order.  *sigh*

So now, I think we'll need to add attlognum in 7.4 and tell admin writers to
change their ORDER BY clauses...

Chris



pgsql-hackers by date:

Previous
From: "Christopher Kings-Lynne"
Date:
Subject: Re: PGXLOG variable worthwhile?
Next
From: Bruce Momjian
Date:
Subject: Re: PGXLOG variable worthwhile?