Date:         Fri, 8 Jan 1999 10:00:45 -0600
Reply-To:     Jack Hamilton <jack_hamilton@FIRSTHEALTH.COM>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         Jack Hamilton <jack_hamilton@FIRSTHEALTH.COM>
Subject:      Re[2]: Should this field get indexed?
Comments: To: jonesj@PHARMARESEARCH.COM
Content-Type: text/plain; charset=US-ASCII

"John T. Jones" <jonesj@PHARMARESEARCH.COM> wrote:
 
>Try it.  SAS uses an algorithm to determine whether
>to use the index or not.  Let it decide.
 
I don't get the impression that it's very smart.  You can use the
$IWEIGHT system option to influence that decision (I've never tried it).
 
I would probably build the index, and use $IWEIGHT to encourage the
system to the index when I expect most of the matches to be on the lower
6-values, and to read sequentially when most of the matches will be
against the 75% value.  But I'd do some experiments before I committed
myself to that.
 
Indexing is supposed to get a lot better in V7 - SAS will keep track of
the distribution values, etc.
 
>There's an option that will put a message to the log
>letting you know whether the index was used or not.
>Sorry, can't recall what it is.
 
   options msglevel=i;
 
==========
>>Dear SAS-L,
>>
>>   We have a 100M row table with a column which takes on only seven distinct
>>values.  On of the values accounts for about 75% of the claims.  This column
>>is an important dimension in many applications.  Does it make sense to build
>>an index on a column with these characteristics?  TIA - Jack