>>> On Fri, Jan 4, 2008 at  6:46 PM, in message <25278(dot)1199493975(at)sss(dot)pgh(dot)pa(dot)us>,
Tom Lane <tgl(at)sss(dot)pgh(dot)pa(dot)us> wrote: 
 
> 8.2 (at least at branch tip, can't say for sure about earlier
> dot-releases)
 
8.2.4 and 8.2.5 both behave this way.
 
>  f2 | f3 | f1 
> ----+----+----
>  53 |    | 53
> (1 row)
> 
> which I claim is the wrong answer.
 
I agree that it is the wrong answer.
 
> the first question for you is whether it is intentional that your query
> joins CTHE to CHST rather than to CH.
 
"CTHE"."eventType" has to match to CHST; there is not an eventType
column in CH.  The other values could refer to either; I would expect
that to be logically equivalent (unless I'm missing something).
 
-Kevin
 

In response to

• Re: OUTER JOIN performance regression remains in 8.3beta4 at 2008-01-05 00:46:15 from Tom Lane

pgsql-hackers by date

Next:	From: Gregory Stark	Date: 2008-01-07 16:15:35
	Subject: Re: Index trouble with 8.3b4
Previous:	From: Andrew Sullivan	Date: 2008-01-07 15:57:53
	Subject: Re: VACUUM FULL out of memory

pgsql-patches by date

Next:	From: Magnus Hagander	Date: 2008-01-08 14:01:53
	Subject: Re: win32.mak patch
Previous:	From: Kevin Grittner	Date: 2008-01-07 15:01:50
	Subject: Re: OUTER JOIN performance regression remains in8.3beta4