Oracle probably created the histogram because
there were no statistics to tell it not to and no clues
about prior use of the columns. For small test
cases, the internal code seems to run a test query
to generate a "height-balanced" histogram - and
then decide not to store it. However, in the case
of the 5 distinct values, Oracle determines that
the number of distinct values is small and also
runs the code for a "value based" histogram, and
stores the results.
--
Regards
Jonathan Lewis
http://www.jlcomp.demon.co.uk/faq/ind_faq.html
The Co-operative Oracle Users' FAQ
http://www.jlcomp.demon.co.uk/seminar.html
Optimising Oracle Seminar - schedule updated July 27th
"Oradba Linux" <techiey2k3_at_comcast.net> wrote in message
news:sHbXc.184007$8_6.108589_at_attbi_s04...
> Using DBMS_STATS to estimate statistics on a table . I created a table
with
> 50,000 rows and it has one index . That column
> has 5 distinct values . I ran a command to estimate statistics on that
table
> with skewonly option . It still went ahead and
> created histograms . I was under the impression that since data
distribution
> is not skewed it will not create histograms .
> Secondly , how does it know that data distribution is skewed or not skewed
> using column stats generated .
>
> I also want an opinion on how to decide whether to gather column stats for
> all columns or just for indexed columns .
> How to decide ?
>
>
>
Received on Sun Aug 29 2004 - 06:57:20 CDT
