From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
---|---|
To: | "Kevin Grittner" <Kevin(dot)Grittner(at)wicourts(dot)gov> |
Cc: | pgsql-performance(at)postgresql(dot)org |
Subject: | Re: LIKE without wildcard different from = |
Date: | 2010-08-04 16:41:47 |
Message-ID: | 21079.1280940107@sss.pgh.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-performance |
"Kevin Grittner" <Kevin(dot)Grittner(at)wicourts(dot)gov> writes:
> We have been using the C locale for everything at our site, but
> there is occasionally talk of supporting characters outside the
> ASCII7 set. In playing around with indexing, to see what the impact
> of that would be, I stumbled across something which was mildly
> surprising.
> In the C locale, if you want to search for an exact value which
> doesn't contain wildcard characters, it doesn't matter whether you
> use the 'LIKE' operator or the '=' operator. With LATIN1 encoding,
> it made three orders of magnitude difference, both in the estimated
> cost and the actual run time.
What PG version are you testing? 8.4 and up should know that an
exact-match pattern can be optimized regardless of the lc_collate
setting.
regards, tom lane
From | Date | Subject | |
---|---|---|---|
Next Message | Kevin Grittner | 2010-08-04 16:52:24 | Re: LIKE without wildcard different from = |
Previous Message | Kevin Grittner | 2010-08-04 16:34:12 | LIKE without wildcard different from = |