Re: re "CBO - A Configuration Roadmap" -- Histograms on Non-Indexed Columns

Christian, Wolfgang, Mladen,

I knew about cardinality driving execution plan if it is used to determine the access method against the table. However, if there was no index on the column, the only access on that column _standalone_ would be a full table scan. {exactly the first paragraph's in the section on Histograms on Non-Indexed columns in Wolfgang's paper , eh ?|
Now, from your emails and Christian's short example and Wolfgang's paper, I take the explanation that this cardinality would be used to determine the JOIN method (sort-merge , nested-loop or hash).

Thank you.

[[BTW : I read Christian's paper after encountering an issue with a query on a single table [no join] where a column in the query IS quite skewed but I thought that Christian's statement was a generalization and began wondering how the Histogram would help in my query {single table , no join}. I guess I was too quick to make that assumption. ]]

So, now, I should also consider skewness in a column if it is in a Join.

Hemant K Chitale
http://web.singnet.com.sg/~hkchital

--
http://www.freelists.org/webpage/oracle-l