Date:         Tue, 10 Nov 1998 15:11:41 -0500
Reply-To:     "Ashiru, Babatunde" <AshiruB@WHITEOAKSEMI.COM>
Sender:       "SAS(r) Discussion" <SAS-L@UGA.CC.UGA.EDU>
From:         "Ashiru, Babatunde" <AshiruB@WHITEOAKSEMI.COM>
Subject:      Re: Funky Stuff in SAS?
Comments: To: "Self, Karsten" <Karsten.Self@Schwab.COM>

Ok, Karsten.  I agree with you on the notion that "sorting" is probably not
necessary or should be postponed.  That tactic is actually used in my
instance, i.e., build the database over a 24 hour period with a process that
picks up new input files, from an input queue, at a regular one hour
interval  -- you know the rest of the unix stuff.  However, this process
does not discriminate against duplicate records hence enters the big-bad
PROC SORT NODUPKEY.  I know, I know .... NODUPKEY is not entirely a panacea
to resolving record duplication problems but it's better than some
alternatives.
 
Proc Sort on the Master dataset is only invoked once a day because a) to
remove duplicates, b) to update another database in a public folder ... it
might be of interest to know that a composite index key was actually defined
to speed things up ... I am not so sure if this index causes more problems
than it actually solves.  I think we may have to purchase the SAS/SPDS to
perhaps permanently resolve this issue.  Not sure.
 
BRA
 
        -----Original Message-----
        From:   Self, Karsten [SMTP:Karsten.Self@Schwab.COM]
        Sent:   Tuesday, November 10, 1998 2:50 PM
        To:     SAS-L@UGA.CC.UGA.EDU; 'Ashiru, Babatunde'
        Subject:        RE: Funky Stuff in SAS?
 
        My usual question is whether or not the sort is necessary.  It often
        isn't, or can be postponed, if you can do subsetting with a format,
        apply CLASS variables to a summary, or use an index.  A standard
        practice is: