Archives of the TeradataForum
Message Posted: Thu, 26 Aug 2004 @ 20:47:08 GMT
A feature. Definitly a feature. The parser didn't use to look at the value the USER function returns. The parser gets the value of USER from the using data parcel. (There is always a using data parcel even when there is no using clause in the request. It contains things like USER, DATE, DATABASE, and so on.)
It used to be the case that the parser did not look at any of the values in the using data parcel because of concerns regarding cached plans, but - sometime during V2R3 I think - we decided it was the lesser of two evils to look at the values like USER. The result is more accurate selectivity estimates for terms like:
<column-reference> = USER
when you have collected statistics on <column-reference>.
I'm sorry to hear the more accurate estimate is not helping you get an LTST plan. Anything that makes the dimension tables look smaller will probably advance your cause.
|Copyright 2016 - All Rights Reserved|
|Last Modified: 27 Dec 2016|