| From: | Josh Berkus <josh(at)agliodbs(dot)com> |
|---|---|
| To: | pgsql-hackers(at)postgresql(dot)org |
| Cc: | chadzakary(at)hotmail(dot)com |
| Subject: | Re: [DEFAULT] Daily digest v1.4318 (23 messages) |
| Date: | 2004-03-10 17:21:54 |
| Message-ID: | 200403100921.54188.josh@agliodbs.com |
| Views: | Whole Thread | Raw Message | Download mbox | Resend email |
| Thread: | |
| Lists: | pgsql-hackers |
Chad,
> I'm talking about the stuff that the other poster (cant see his name
> right now, sorry) doubts will ever be in postgres. ie you can seek to
> anywhere in the Btree using a row offset as a "search" key.
And this is more useful than LIMIT # OFFSET # on queries, how, exactly?
> I'd love to hear why this would be hard to support in a materialized
> view. Could you explain that ? Berkeley DB supports it.
Berkeley DB is not a Relational Database.
It's not a question of "hard to support". It's a question of "don't want to
support". One of the core tenets of relational database theory is that
there are no row numbers; rows only have a fixed order as a part of a sorted
final output set (e.g. a query with an ORDER BY).
I don't know what kind of application you're trying to support that you think
row numbers are such a keen idea. As far as we're concerned, row numbers
are an inefficient throwback to the pre-relational databases of the early
1980's; why would we want them?
--
-Josh Berkus
Aglio Database Solutions
San Francisco
| From | Date | Subject | |
|---|---|---|---|
| Next Message | Andrew Dunstan | 2004-03-10 17:23:38 | selective statement logging |
| Previous Message | Merlin Moncure | 2004-03-10 17:16:06 | Re: optimizing impossible matches |