From: | "Ron Peacetree" <rjpeace(at)earthlink(dot)net> |
---|---|
To: | pgsql-hackers(at)postgresql(dot)org |
Subject: | Re: No merge sort? |
Date: | 2003-04-08 02:11:07 |
Message-ID: | %cqka.13403$ey1.1154620@newsread1.prod.itd.earthlink.net |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
<cbbrowne(at)cbbrowne(dot)com> wrote in message
news:20030407202001(dot)1EC3C58E0D(at)cbbrowne(dot)com(dot)(dot)(dot)
> It is highly likely that it will typically take more
> computational effort to figure out that one of the 4 sorts provided
> /any/ improvement than any computational effort they would save.
>
> That's a /very/ common problem. There's also a fair chance, seen in
> practice, that the action of collecting additional statistics to
improve
> query optimization will cost more than the savings provided by the
> optimizations.
>
"Back in the Day" I heard similar arguments when discussing whether
there should be support for hashing [O(n)], interpolation search
[O(lglg(n))], binary search [O(lg(n))], and sequential search [O(n)]
or for only some subset of these for a DB system I was working on. To
make a long story short, it was worth it to have support for all of
them because the "useful domain" of each was reasonably large and
reasonably unique compared to the others.
I submit a similar situation exists for sorting. (and yes, I've been
here before too ;-)
Giving end users of PostgreSQL a reasonable set of tools for building
high performance applications is just good business.
From | Date | Subject | |
---|---|---|---|
Next Message | Stephan Szabo | 2003-04-08 02:13:24 | Re: Backpatch FK changes to 7.3 and 7.2? |
Previous Message | Brian E. Pangburn | 2003-04-08 02:01:51 | A funny story about horology regression test failures... |