hi tom ...
the idea behind this is to disable the stats on a certain column
entirely.
this would give me more control about the plan. in this special case
data is changing so frequently that the default values are ways
better than trying to keep the "real" stats up to date.
in case of default value i know what the DB does given a certain
where clause - this is beyond my control when stats drop in.
i guess there are corner cases where no stats on certain fields can
definitely help to make plans a little bit more stable.
many thanks,
hans
On Aug 29, 2007, at 6:44 PM, Tom Lane wrote:
Hans-Juergen Schoenig <[EMAIL PROTECTED]> writes:
i came across some interesting behavior of pg_stats and i am not sure
if this is something we should treat the way we do it.
Setting target zero means "expend no work on this column". In my book
that includes not doing anything to any pre-existing pg_stats entry.
What you propose would defeat the ability to analyze an unchanging
column once and then make ANALYZE skip over it henceforth.
regards, tom lane
---------------------------(end of
broadcast)---------------------------
TIP 3: Have you checked our extensive FAQ?
http://www.postgresql.org/docs/faq
--
Cybertec Geschwinde & Schönig GmbH
Gröhrmühlgasse 26, 2700 Wiener Neustadt
Tel: +43/1/205 10 35 / 340
www.postgresql.at, www.cybertec.at
