From: | Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> |
---|---|
To: | "Baegle" <al(at)wifiny(dot)org> |
Cc: | pgsql-bugs(at)postgresql(dot)org |
Subject: | Re: BUG #5530: Default cost of crypt causes poor decisions |
Date: | 2010-06-29 22:21:31 |
Message-ID: | 25610.1277850091@sss.pgh.pa.us |
Views: | Raw Message | Whole Thread | Download mbox | Resend email |
Thread: | |
Lists: | pgsql-bugs |
"Baegle" <al(at)wifiny(dot)org> writes:
> The default cost of crypt() is 1, but crypt is intentionally a high cost
> function. This mis-match between assigned cost and actual cost causes the
> query planner to use crypt too soon and slows down execution of queries
> significantly. The cost of the function should be more congruent with it's
> intended cost.
Perhaps, but the same could be said of most of the contrib/pgcrypto
functions, no? Which ones do you think should have a nondefault cost,
and what should we set it to?
Also, it'd be easier to persuade people this is worth worrying about if
you showed a real-world case where it's a problem. crypt() doesn't seem
like a function that would be very likely to be used in contexts where
this'd matter.
regards, tom lane
From | Date | Subject | |
---|---|---|---|
Next Message | Bruce Momjian | 2010-06-29 22:24:25 | Re: local_preload_libraries filenames converted to lowercase |
Previous Message | Tom Lane | 2010-06-29 21:21:26 | Re: Function works in 8.4 but not in 9.0 beta2 "ERROR: structure of query does not match function result type" |