From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
---|---|
To: | Robert Haas <robertmhaas(at)gmail(dot)com> |
Cc: | Dan McGee <dan(at)archlinux(dot)org>, pgsql-hackers(at)postgresql(dot)org |
Subject: | Re: [PATCH] Never convert n_distinct < 2 values to a ratio when computing stats |
Date: | 2012-03-25 15:59:22 |
Message-ID: | 15920.1332691162@sss.pgh.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-hackers |
Robert Haas <robertmhaas(at)gmail(dot)com> writes:
> The bit about maybe not getting both t and f as MCVs on a Boolean does
> seem a little worrying, but I'm not sure whether it actually affects
> query planning in a materially negative way. Can you demonstrate a
> case where it matters?
If we were trying to force that to happen it would be wrong anyway.
Consider a column that contains *only* "t", or at least has so few
"f"'s that "f" appears never or only once in the selected sample.
(IIRC there is a clamp that prevents selecting anything as an MCV
unless it appears at least twice in the sample.)
Like Robert, I'm not convinced whether or not this is a reasonable
change, but arguing for it on the basis of boolean columns doesn't
seem very sound.
regards, tom lane
From | Date | Subject | |
---|---|---|---|
Next Message | Andrew Dunstan | 2012-03-25 16:12:33 | occasional startup failures |
Previous Message | Robert Haas | 2012-03-25 15:32:04 | Re: Gsoc2012 Idea --- Social Network database schema |