On Sun, Mar 4, 2012 at 11:22 PM, <maxim(dot)boguk(at)gmail(dot)com> wrote:
> I got hit by that bug when explored reasons of one very slow production
> query again.
> And I lost a lot time trying reproduce the problem query on production
> server with explain analyze.
> Finally I found I need some workaround to get explain perform actual work
> with toasted data.
> So there is going the bug report:
> EXPLAIN do not take into account that some result fields will require
> detoasting and produce wrong buffers result.
Well, the buffers result is, I think, not wrong. That's just a count
of how many buffers were actually read or written or whatever, and it
is what it is. I think that if there's an actual bug here, it's that
EXPLAIN ANALYZE is skipping the formation of the actual output tuples,
and therefore it's doing less work than an actual execution of the
real query would. I am not sure whether it would be a good idea to
change that or not.
The Enterprise PostgreSQL Company
In response to
pgsql-bugs by date
|Next:||From: Tom Lane||Date: 2012-03-12 15:16:18|
|Subject: Re: BUG #6517: Volatile function erroneously optimized, does not consider change in schema path |
|Previous:||From: Robert Haas||Date: 2012-03-12 15:04:21|
|Subject: Re: BUG #6517: Volatile function erroneously optimized, does
not consider change in schema path|